- Email: [email protected]
Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares
Author’s Accepted Manuscript Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares Gustavo G. Shimamoto, Matthieu Tubino www.elsevier.com/locate/talanta
PII: DOI: Reference:
S0039-9140(17)31205-5 https://doi.org/10.1016/j.talanta.2017.12.001 TAL18137
To appear in: Talanta Received date: 11 June 2017 Revised date: 30 November 2017 Accepted date: 1 December 2017 Cite this article as: Gustavo G. Shimamoto and Matthieu Tubino, Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares, Talanta, https://doi.org/10.1016/j.talanta.2017.12.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares Gustavo G. Shimamoto, Matthieu Tubino* Institute of Chemistry, University of Campinas - UNICAMP, P.O. Box 6154, 13083-970, Campinas-SP, Brazil. *E-mail: [email protected].
Abstract: Biodiesel quality is checked by determining several parameters. Considering the large number of analyses in this verification, as well as the disadvantages of the use of toxic solvents and waste generation, multivariate calibration is suggested to reduce the number of tests. In this work, hydrogen nuclear magnetic resonance (1H NMR) spectra were used to build multivariate models, from partial least squares (PLS), in order to perform simultaneous determination of six important quality parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability, and water content. 1H NMR spectrum reflects the structures of the compounds present in biodiesel and showed suitable correlations with the six parameters. In addition, the models were appropriate to predict all parameters for external samples. Thus, the alliance between 1H NMR spectra and PLS was shown to be applicable to extract a lot of information about biodiesel quality, significantly reducing analysis time, reagent and solvent consumption, and waste generation.
Keywords: Biodiesel; Quality parameters; Nuclear magnetic resonance; Chemometrics
1
1. Introduction Biodiesel is an alternative fuel that is added to mineral diesel. This mixture can be used in diesel motors, because biodiesel exhibits similar properties to those of diesel. Additionally, it is obtained from renewable raw materials, and is biodegradable and non-toxic, with less environmental impact regarding emissions [1,2]. Biodiesel is produced through the transesterification of oils or fats using a short chain alcohol in the presence of a catalyst. The product of the transesterification reaction is a mixture of biodiesel (alkyl esters), glycerides that have not been completely transesterified, glycerol (as a byproduct), residual catalyst and residual alcohol. All of these compounds must be removed from biodiesel to ensure a good quality product [3]. Biodiesel quality is checked by determining several parameters. Furthermore, to guarantee good quality, biodiesel parameters have to be in agreement with those specified by standards, such as ASTM D6751-5a (American), BS EN 14214:2012+A1:2014 (European), or ANP Resolution Nº 45, 2014 (Brazilian) [4-6]. In fact, there are more than twenty analyses that must be performed in order to check whether a biodiesel can be marketed or not. The biodiesel parameters that were studied in this work are: density, kinematic viscosity, iodine value, acid number, oxidative stability and water content. The density of the biodiesel is directly related to the chemical structure of the compounds from which it is constituted. The carbon chain size and the number of unsaturations directly influence the density. Larger length of carbon chain of the alkyl ester means higher density;
2
however, density decreases with the increase of the degree of unsaturation. Also, impurities and additives may influence in the biodiesel density [3,7]. Viscosity is also directly related to the molecular structures and is defined as a measure of fluid resistance to flow. Concerning biodiesel kinematic viscosity, it increases with chain length and is strongly dependent on the nature and number of double bonds. Furthermore, it increases sharply with polymerization [8]. Iodine value is a measure of the number of double bonds in a sample. It specifies the mass of iodine (I2) in grams that is consumed by 100 grams of sample. Some factors that influence the iodine value are storage conditions and biodiesel age, especially if the sample has undergone oxidation processes. The oxidation reactions include a series of complex chemical reactions characterized by decreases in the amount of unsaturation in the fatty acid chains and the formation of degradation compounds, such as alcohols, aldehydes, ketones, acids, and other products [9-12]. Acid number indicates the acid products of degradation reactions of oils or biodiesel and is expressed in milligrams of KOH per gram of sample. Therefore, the determination of their concentrations during storage is important for monitoring the occurrence of degradation reactions. These reactions lead to the breakdown of the triglyceride molecule and to the formation of free fatty acids [13]. Oxidative stability is a comparative parameter that is widely applied in the quality control of raw material and is very useful for the evaluation of different types of oils or biodiesel, as well as in determining the efficiency of antioxidant additives. The method used for the determination of oxidative stability is known as Rancimat and involves exposing the sample to an airflow at a
3
high temperature, usually 110 ºC. Under these conditions, the formation of the resulting compounds from oxidation is intensified, and the volatiles are conducted to measure cells containing deionized water, whose conductivity is continuously measured. Oxidative stability is expressed by the induction period (in hours), being determined at the point when the conductivity level, in the collecting flask, increases rapidly, signaling the onset of the overall oxidation process [14]. Water content may promote oil or biodiesel degradation and the proliferation of microorganisms, affecting its quality [15]. Karl Fischer titration is the most frequently used method for measuring the water content (humidity) in oils, solvents and other products. The method involves the titration with Karl Fischer reagent of a sample dissolved in methanol [16]. In this context, considering the large number of analyses, as well as the disadvantages of the use of toxic solvents and waste generation, multivariate calibration is suggested to reduce the number of tests. Therefore, in this work, hydrogen nuclear magnetic resonance (1H NMR) spectra were used to build multivariate models, from partial least squares (PLS), in order to perform simultaneous determination of six important quality parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability and water content. From 1H NMR, it is possible to extract a lot of informations about organic compounds. Despite the considerable cost of a NMR spectrometer, there are many advantages to using this technique. For example: qualitative/quantitative analysis can be done simultaneously; measurements can be performed quickly (for an abundant nucleus, such as 1H); it is not required to isolate the analyte when present in mixtures; analysis can be performed simultaneously for different analytes in a
4
single spectrum [17]. Furthermore, there is often NMR equipment for multiple users available in many institutions.” 2. Materials and methods 2.1. Biodiesel Samples Biodiesels from twelve different vegetable oil sources were used, such as babassu, brown flaxseed, canola, corn, cottonseed, macauba kernel, microalgae, palm, residual frying, sesame, soybean, and sunflower. Refined oils were employed for biodiesel synthesis without further purification. For the unrefined oils and the residual frying oil, because of their high acid value, it was necessary to perform acid esterification before transesterification in order to decrease the quantity of free fatty acids. The transesterification reaction was performed using the optimized procedure as previously reported, using analytical grade methanol as the alcohol and sodium methoxide, 30% w/w in methanol, as the catalyst [18]. 2.2. 1H NMR Spectra All of the 1H NMR spectra were recorded in a Bruker Avance III 500 MHz NMR spectrometer. To obtain the spectra, a sample of 20 µL of each biodiesel was dissolved in 600 µL of CDCl3, containing tetramethylsilane (TMS) as an internal reference (0.00 ppm), using the following experimental conditions: pulse program, zg30; spectral width, from –4.00 to 16.00 ppm; spectral size, 32768 points; 90º pulse, 11.75 µs; delay, 5 s; and number of scans, 16. The spectra were obtained in triplicate for each sample, and only 2 minutes was taken to obtain each spectrum. 2.3. Density at 20 ºC
5
Slightly less than 5 mL of the sample was added to a calibrated 5 mL volumetric flask of known weight. The flask with the sample was thermostated in a water bath at 20.0 ± 0.1 ºC for 10 minutes. Then, the flask was brought to volume at this temperature. The flask was dried with paper towels and weighed. The density was calculated based on the mass/volume relationship. 2.4. Kinematic Viscosity at 40 ºC by ASTM D445 The kinematic viscosity was determined according to ASTM D445 [19], using glass kinematic viscometer tubes Cannon Fenske and a thermostated bath Quimis. 2.5. Iodine Value by Wijs Method (BS EN 14111) The iodine value was determined according to BS EN 14111, using a 50 mL digital manual burette [20]. 2.6. Acid Number by AOCS Cd 3d-63 The acid number was determined according to AOCS Cd 3d-63 [21]. To perform the titration, a Metrohm Titrando 809 was used with a Metrohm Solvotrode electrode (electrolytesaturated LiCl in ethanol) and an automatic sampler Metrohm 814 USB Sample Processor. 2.7. Oxidative Stability at 110 ºC and under 10 L h-1 air flow by Rancimat (BS EN 14112) The induction period, which expresses the oxidative stability, was determined through the BS EN 14112 method using a Metrohm Rancimat model 873 [22].
2.8. Water Content by Coulometric Karl Fischer Titration (BS EN ISO 12937)
6
The total water content in biodiesel was determined according to BS EN ISO 12937 [23], using Karl Fischer coulometric titration (Metrohm model 831). 2.9. Chemometric Procedures by Partial Least Squares (PLS) PLS was performed using the Pirouette® 4.5 software. First, as pretreatment, each 1
H NMR spectrum was normalized from 0 to 1, dividing all points of each spectrum by its
highest value, to mitigate the influence of the biodiesel concentration in the solvent (CDCl3). Then, the first derivative of each spectrum was obtained to avoid influences of baseline variations. Finally, as preprocessing, the spectra were mean centered. This procedure involves the subtraction of the element of each column (variable) by the average value of the elements of that column (variable), the result is a matrix in which all columns (variables) have zero mean. Besides this preprocessing, the variables related to the solvent and TMS were excluded to perform the procedures since these spectral regions do not bring information related to the samples. Therefore, it was employed spectral width from 0.50 to 5.70 ppm (16908 points).
3. Results and discussion Table 1 shows the results obtained for the six quality parameters of the biodiesel samples. The different types of esters present in biodiesel, which generate different quality parameters, could be differentiated by the 1H NMR spectra. These differences are noted mainly by the number of unsaturations, the positions, and the length of the ester carbon chains. [Table 1]
7
The 1H NMR spectra obtained for all biodiesel samples are shown overlaid in Fig. 1 to illustrate the type of information that was used to build the multivariate models, using the PLS chemometric tool. [Fig. 1] The PLS regression method is a multivariate calibration procedure that uses Principal Component Analysis (PCA) to reduce the size of the data set and a correlation is made between this data set, such as spectra, and properties of interest, wherein these properties of interest can be physical or chemical characteristics. As in the calculation of principal components (PC), the property values of interest are taken into account (supervised method), the PC are named latent variables (LV) [24]. Figure 2 shows the graphs for reference values versus predicted values by calibration and by cross-validation. [Fig. 2] In Fig. 2, appropriate calibration can be noted, with correlation coefficients of calibration (rcal) being higher than 0.99, and suitable cross-validation with correlation coefficients of crossvalidation (rcv) being higher than 0.98. This indicates accentuated correlations between the 1
H NMR spectra and all of the studied parameters. In Table 2, the external validation results for the six quality parameters are presented as
the average of three replicates. Validations were performed considering each type of biodiesel, not each replicate. In order to calculate the relative errors, the mean measured values were considered as the true values.
8
[Table 2] The highest mean relative error value in the external validation was 12% for acid number; however, this parameter showed very low values. For instance, the maximum absolute error between the reference value and the correspondent predicted value is about 0.04 mg KOH g-1 that indeed is quite low for this attribute. In the other cases, the mean relative errors are low enough to allow the use in the determination of the related proprieties. Therefore, the PLS models proposed above can be recommended for the screening analysis of biodiesel concerning six quality parameters by linking 1H NMR spectra and chemometrics. In general, screening analysis is related to methods that normally involve little or no sample treatment, with the fast acquisition of qualitative, semi-quantitative, or even quantitative data about many characteristics of the sample; their response is used for decision-making, and oftentimes, confirmation by conventional methods is unnecessary [25]. The multivariate model accuracy is expressed by root mean square error (RMSE) of calibration (RMSEC), of cross-validation (RMSECV), obtained by internal validation, and of prediction (RMSEP), which is calculated from an external validation set of samples. Equation 1 represents the general equation of errors: n
(y yˆ ) i
RMSE
2
i
i1
n
Where: n = the number of samples; ̂ = predicted value for i-th sample;
(1) = reference
value for i-th sample. All of these kinds of errors are presented in Table 3, where the correlation coefficients of calibration (rcal), validation (rcv), number of latent variables (LV), and working range for all
9
proposed multivariate models are shown. The minimum number of LV was chosen based on RMSECV values, calculated for each LV. As shown, for example, in Figure 3, RMSECV values are almost constant from eighth LV for the PLS model that predicts acid number. For the other models, the number of LV was chosen analogously. [Table 3] [Figure 3] The parameters in Table 3 confirm the suitable calibrations, internal validations and predictive power of the chemometric models. As a result, as well as determining the molecular structures of organic compounds, the 1H NMR spectrum can provide information about several quality parameters of its compounds when chemometric tools are applied. From the results obtained (Table 2 and 3), it is possible to observe a better performance of the models to determine viscosity, density, and iodine value, because they showed suitable mean relative differences (Table 2) and it was not necessary to exclude samples to build the models, so that the entire working range among the samples was used. These parameters are more directly related to the major compounds present in biodiesel (methyl esters); because of this, they were probably easier to calibrate, validate, and predict. On the other hand, to build the models for acid number, induction period, and water content, some samples must be excluded to get an appropriate adjustment, because they presented much higher or much lower parameters compared with the other samples. In addition, the induction period is a parameter that depends not only on the structure of the methyl esters. Impurities from the synthesis, for example, alcohol, or the presence of metal ions or other organic compounds may influence the oxidative stability, even at low concentrations [26]. Fatty
10
acids can be detected and differentiated from the methyl esters by 1H NMR spectra [17]; however, they are present in biodiesel at very low concentrations. In this context, it is believed that the models to determine induction period and acid number showed higher mean relative errors for these reasons (Table 2). From these results, the application of the developed PLS models is proposed for the screening analysis of biodiesel parameters. These models allow the simultaneous analysis of six parameters, reducing analysis time, reagent consumption and therefore waste generation. Some multivariate models have been built for the determination of biodiesel quality parameters, such as density, water content, sulfur content, oxidative stability, and residual alcohol, among others [27-31]. However, all of them use infrared spectroscopy as the source of data. In this context, the development of other kinds of multivariate models, from 1H NMR spectroscopy, is interesting as they offer more analytical possibilities.
4. Conclusion This study demonstrated correlations between 1H NMR data with six studied parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability and water content. Therefore, a screening analysis to enable the simultaneous determination of these six quality parameters has been developed from 1H NMR and PLS. The method showed good correlation, accuracy, and low relative errors. Once the multivariate model is built and implemented, it determines at least six important parameters of biodiesels, using simple and rapid procedures. Also, it minimizes experimental steps, analysis time, the consumption of reagents and solvents and waste generation.
11
Acknowledgements The authors are grateful to the Brazilian National Council of Technological and Scientific Development (CNPq) for financial support.
References [1] A.E. Atabani, A.S. Silitonga, I.A. Badruddin, T.M.I. Mahlia, H.H. Masjuki, A.S. Mekhilef, A comprehensive review on biodiesel as an alternative energy resource and its characteristics, Renewable Sustainable Energy Rev. 16 (2012) 2070-2093. [2] D.Y.C. Leung, B.C.P. Koo, Y. Guo, Degradation of biodiesel under different storage conditions, Bioresour. Technol. 97 (2006) 250-256. [3] I.P. Lôbo, S.L.C. Ferreira, R.S. Cruz, Biodiesel: parâmetros de qualidade e métodos analíticos, Quim. Nova 32 (2009) 1596-1608. [4] ASTM D6751 – 15a, Standard specification for biodiesel fuel blend stock (B100) for middle distillate fuels, 2015. [5] BS EN 14214:2012+A1:2014, Liquid petroleum products – fatty acid methyl esters (FAME) for use in diesel engines and heating applications – requirements and test methods, 2014. [6] Agência Nacional do Petróleo, Gás Natural e Biocombustíveis (ANP), Resolução Nº 45, 2014.
12
[7] B. Tesfa, R. Mishra, F. Gu, N. Powles, Prediction models for density and viscosity of biodiesel and their effects on fuel supply system in CI engines, Renewable Energy 35 (2010) 2752-2760. [8] G. Knothe, K.R. Steidley, Kinematic viscosity of biodiesel fuel components and related compounds. Influence of compound structure and comparison to petrodiesel fuel components, Fuel 84 (2005) 1059-1065. [9] G.G. Shimamoto, M.M.A. Favaro, M. Tubino, Simple methods via mid-IR or 1H NMR spectroscopy for the determination of the iodine value of vegetable oils, J. Braz. Chem. Soc. 26 (2015) 1431-1437. [10] G. Knothe, Structure indices in FA chemistry. How relevant is the iodine value?, J. Am. Oil Chem. Soc. 79 (2002) 847-854. [11] G.G. Shimamoto, J.A. Aricetti, M. Tubino, A simple, fast, and green titrimetric method for the determination of the iodine value of vegetable oils without Wijs solution (ICl), Food Anal. Methods 9 (2016) 2479–2483. [12] M. Tubino, J.A. Aricetti, A green potentiometric method for the determination of the iodine number of biodiesel, Fuel 103 (2013) 1158-1163. [13] M. Tubino, J.A. Aricetti, A visual titration method for the determination of the acid number of oils and fats: A green alternative, J. Am. Oil Chem. Soc. 89 (2012) 2113-2115. [14] J. Pullen, K. Saeed, An overview of biodiesel oxidation stability, Renewable Sustainable Energy Rev. 16 (2012) 5924-5950.
13
[15] G. Knothe, Some aspects of biodiesel oxidative stability, Fuel Process. Technol. 88 (2007) 669-677. [16] A. Felgner, R. Schlink, P. Kirschenbühler, B. Faas, H.D. Isengard, Automated Karl Fischer titration for liquid samples – Water determination in edible oils, Food Chem. 106 (2008) 13791384. [17] J.H. Simpson, Organic structure determination: using 2-D NMR spectroscopy: a problembased approach, Elsevier/Academic, Amsterdam, 2008. [18] W.L.G Silva, P.T. Souza, G.G. Shimamoto, M. Tubino, Separation of the glycerol-biodiesel phases in an ethyl transesterification synthetic route using water, J. Braz. Chem. Soc. 26 (2015) 1745-1750. [19] ASTM D445–12, Standard test method for kinematic viscosity of transparent and opaque liquids (and calculation of dynamic viscosity), 2012. [20] BS EN 14111, Fat and oil derivatives – fatty acid methyl esters (FAME) – determination of iodine value, 2003. [21] American Oil Chemists' Society, AOCS Official Method Cd 8d-63, in: D. Firestone (Ed), Official methods and recommended practices of the AOCS, AOCS, Champaign, 1973. [22] BS EN 14112, Fat and oil derivatives. Fatty acid methyl esters (FAME). Determination of oxidation stability (accelerated oxidation test), 2003. [23] BS EN ISO 12937, Methods of test for petroleum and its products. BS 2000-438. Petroleum products. Determination of content. Coulometric Karl Fischer titration method, 2001.
14
[24] K.R. Beebe, R.J. Pell, M.B. Seasholtz, Chemometrics: a practical guide, John Wiley & Sons, New York, 1998. [25] R. Muñoz-Olivas, Screening analysis: an overview of methods applied to environmental, clinical and food analyses, Trends Anal. Chem. 23 (2004) 203-216. [26] N. Canha, P. Felizardo, J.C. Menezes, J.N. Correia, Multivariate near infrared spectroscopy models for predicting the oxidativestability of biodiesel: Effect of antioxidants addition, Fuel 97 (2012) 352-357. [27] M. Bampi, A.P. Scheer, F. Castilhos, Application of near infrared spectroscopy to predict the average droplet size and water content in biodiesel emulsions, Fuel 113 (2013) 546-552. [28] P. Felizardo, P. Baptista, J.C. Menezes, M.J.N. Correia, Multivariate near infrared spectroscopy models for predicting methanol and water content in biodiesel, Anal. Chim. Acta 595 (2007) 107-113. [29] R.M. Balabin, R.Z. Safieva, Near-infrared (NIR) spectroscopy for biodiesel analysis: fractional composition, iodine value, and cold filter plugging point from one vibrational spectrum, Energy Fuels 25 (2011) 2373-2382. [30] L.F.B. Lira, M.S. Albuquerque, J.G.A. Pacheco, T.M. Fonseca, E.H.S. Cavalcanti, L. Stragevitch, M.F. Pimentel, Infrared spectroscopy and multivariate calibration to monitor stability quality parameters of biodiesel, Microchem. J. 96 (2010) 126-131.
15
[31] L.F.B. Lira, F.V.C. Vasconcelos, C.F. Pereira, A.P.S. Paim, L. Stragevitch, M.F. Pimentel, Prediction of properties of diesel/biodiesel blends by infrared spectroscopy and multivariate calibration, Fuel 89 (2010) 405-409.
16
Figure Captions Fig. 1. Overlapping 1H NMR spectra of biodiesel samples.
Fig. 2. Relationship between the reference and the predicted values obtained by calibration and by cross-validation via PLS model for density, kinematic viscosity, iodine value, acid number, induction period and water content of biodiesel samples.
Fig. 3. Root mean square error of cross-validation (RMSECV) as function of the number of latent variables (LV) for the PLS model that predicts acid number.
17
Tables Table 1. Quality parameters of biodiesel samples (average of three replicates). Kinematic Viscosity at 40 ºC
Density at 20 ºC
Biodiesel
/ (kg m-3)
/ (mm2 s-1)
Iodine Value
Acid Number
/ (g I2 100 g-1)
/ (mg KOH g-1)
Induction Period
Water Content
/ (h)
/ (mg kg-1)
Babassu
868.13 0.07
2.713 0.006
15.2 0.4
0.670 0.008
10.2 0.3
178 6
Brown flaxseed
884.36 0.03
3.61 0.04
151 6
0.117 0.003
1.91 0.03
193 6
878.0 0.2
5.43 0.04
106.2 0.6
0.22 0.02
9.2 0.3
270 20
Corn
879.08 0.01
4.45 0.03
114.6 0.4
0.188 0.008
7.10 0.03
277 7
Cottonseed
879.25 0.06
4.510 0.009
109 1
0.16 0.01
3.06 0.06
620 30
869.4 0.2
3.003 0.004
30 2
0.204 0.008
61 6
190 20
Microalgae
878.20 0.03
4.660 0.006
107 1
0.15 0.01
7.62 0.09
198 8
Palm
873.18 0.03
3.44 0.02
20.4 0.4
0.240 0.003
55 3
394 5
Residual frying
888.0 0.2
4.616 0.007
113.5 0.5
0.17 0.01
1.61 0.02
390 10
Sesame
879.1 0.1
4.240 0.001
104.9 0.2
0.11 0.01
8.7 0.1
80 20
Soybean
883.6 0.1
4.38 0.01
124.3 0.9
0.201 0.008
6.2 0.4
249 4
Sunflower
885.9 0.1
4.718 0.008
123.4 0.5
0.222 0.002
2.1 0.1
233 6
Canola
Macauba kernel
Table 2. Predicted and measured parameters values for external samples (averages of three replicates) and relative errors. Density at 20 ºC / (kg m-3) Biodiesel
Kinematic Viscosity at 40 ºC / (mm2 s-1)
Iodine Value / (g I2 100 g-1)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Corn
881.55
879.08
0.3
4.63
4.45
4.0
114.9
114.6
0.3
Soybean
887.3
883.6
0.4
4.35
4.38
0.7
121.2
124.3
2.5
Sunflower
884.4
885.9
0.2
4.378
4.718
7.2
127.3
123.4
3.1
Mean Relative Error (%)
Mean
0.3
Mean
Acid Number / (mg KOH g-1) Biodiesel
4.0
Mean
Water Content / (mg kg-1)
2.0
Induction Period / (h)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Corn
0.196
0.188
4.3
276
277
0.4
7.3
7.1
2.8
Soybean
0.181
0.222
18
255
249
2.4
5.9
6.2
4.8
Sunflower
0.171
0.201
15
237
233
1.7
2.36
2.14
11
Mean Relative Error (%)
Mean
12
Mean
1.5
Mean
6.2
18
Table 3. Parameters of PLS models to determine density, kinematic viscosity, iodine value, acid number, induction period and water content. Partial Least Squares (PLS) Models Iodine Value
Acid Number a
Induction Period b
Water Content c
868.13 – 887.97 (kg m-3)
Kinematic Viscosity at 40ºC 2.713 – 5.43 (mm2 s-1)
15.2 – 151 (g I2 100 g-1)
0.11 – 0.240 (mg KOH g-1)
1.61 – 9.2 (h)
178 – 394 (mg kg-1)
12
12
12
11
9
10
36
36
36
33
27
30
16908
16908
16908
16908
16908
16908
Parameter Density at 20ºC Working Range Types of Biodiesel (n=3 replicates) Samples Number of Variables (from 0.5 to 5.7 ppm) Latent Variable
8
7
-3
RMSEC
0.4 (kg m )
rcal
0.9984
RMSECV
0.6 (kg m-3)
rcv
0.9955
RMSEP
3 (kg m-3)
4 8 6 8 3 0.003 0.08 (mm s ) 0.1 (h) 6 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) 0.9958 0.9985 0.9978 0.9994 0.9974 3 0.004 2 -1 0.09 (mm s ) 0.6 (h) 12 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) 0.9930 0.9976 0.9956 0.9835 0.9871 3 0.03 2 -1 0.2 (mm s ) 0.2 (h) 8 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) excluded to build the model, because its acid number is much higher compared to 2 -1
a = babassu biodiesel sample was others. b = babassu, macauba, and palm biodiesel samples were excluded to build the model, because their induction period values are much higher compared to others. c = cottonseed and sesame biodiesel samples were excluded to build the model, because their water content values are much different compared to others.
19
20
Highlights
1
The models can perform simultaneous quantitative analysis of six biodiesel
H NMR spectra were used to build multivariate models
parameters
Parameters: density, viscosity, iodine value, acid number, oxidative stability, water.
There is significant reduction of analysis time, reagents consumption and waste.
21
PII: DOI: Reference:
S0039-9140(17)31205-5 https://doi.org/10.1016/j.talanta.2017.12.001 TAL18137
To appear in: Talanta Received date: 11 June 2017 Revised date: 30 November 2017 Accepted date: 1 December 2017 Cite this article as: Gustavo G. Shimamoto and Matthieu Tubino, Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares, Talanta, https://doi.org/10.1016/j.talanta.2017.12.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Simultaneous determination of six quality parameters of biodiesel through 1H NMR spectroscopy and partial least squares Gustavo G. Shimamoto, Matthieu Tubino* Institute of Chemistry, University of Campinas - UNICAMP, P.O. Box 6154, 13083-970, Campinas-SP, Brazil. *E-mail: [email protected].
Abstract: Biodiesel quality is checked by determining several parameters. Considering the large number of analyses in this verification, as well as the disadvantages of the use of toxic solvents and waste generation, multivariate calibration is suggested to reduce the number of tests. In this work, hydrogen nuclear magnetic resonance (1H NMR) spectra were used to build multivariate models, from partial least squares (PLS), in order to perform simultaneous determination of six important quality parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability, and water content. 1H NMR spectrum reflects the structures of the compounds present in biodiesel and showed suitable correlations with the six parameters. In addition, the models were appropriate to predict all parameters for external samples. Thus, the alliance between 1H NMR spectra and PLS was shown to be applicable to extract a lot of information about biodiesel quality, significantly reducing analysis time, reagent and solvent consumption, and waste generation.
Keywords: Biodiesel; Quality parameters; Nuclear magnetic resonance; Chemometrics
1
1. Introduction Biodiesel is an alternative fuel that is added to mineral diesel. This mixture can be used in diesel motors, because biodiesel exhibits similar properties to those of diesel. Additionally, it is obtained from renewable raw materials, and is biodegradable and non-toxic, with less environmental impact regarding emissions [1,2]. Biodiesel is produced through the transesterification of oils or fats using a short chain alcohol in the presence of a catalyst. The product of the transesterification reaction is a mixture of biodiesel (alkyl esters), glycerides that have not been completely transesterified, glycerol (as a byproduct), residual catalyst and residual alcohol. All of these compounds must be removed from biodiesel to ensure a good quality product [3]. Biodiesel quality is checked by determining several parameters. Furthermore, to guarantee good quality, biodiesel parameters have to be in agreement with those specified by standards, such as ASTM D6751-5a (American), BS EN 14214:2012+A1:2014 (European), or ANP Resolution Nº 45, 2014 (Brazilian) [4-6]. In fact, there are more than twenty analyses that must be performed in order to check whether a biodiesel can be marketed or not. The biodiesel parameters that were studied in this work are: density, kinematic viscosity, iodine value, acid number, oxidative stability and water content. The density of the biodiesel is directly related to the chemical structure of the compounds from which it is constituted. The carbon chain size and the number of unsaturations directly influence the density. Larger length of carbon chain of the alkyl ester means higher density;
2
however, density decreases with the increase of the degree of unsaturation. Also, impurities and additives may influence in the biodiesel density [3,7]. Viscosity is also directly related to the molecular structures and is defined as a measure of fluid resistance to flow. Concerning biodiesel kinematic viscosity, it increases with chain length and is strongly dependent on the nature and number of double bonds. Furthermore, it increases sharply with polymerization [8]. Iodine value is a measure of the number of double bonds in a sample. It specifies the mass of iodine (I2) in grams that is consumed by 100 grams of sample. Some factors that influence the iodine value are storage conditions and biodiesel age, especially if the sample has undergone oxidation processes. The oxidation reactions include a series of complex chemical reactions characterized by decreases in the amount of unsaturation in the fatty acid chains and the formation of degradation compounds, such as alcohols, aldehydes, ketones, acids, and other products [9-12]. Acid number indicates the acid products of degradation reactions of oils or biodiesel and is expressed in milligrams of KOH per gram of sample. Therefore, the determination of their concentrations during storage is important for monitoring the occurrence of degradation reactions. These reactions lead to the breakdown of the triglyceride molecule and to the formation of free fatty acids [13]. Oxidative stability is a comparative parameter that is widely applied in the quality control of raw material and is very useful for the evaluation of different types of oils or biodiesel, as well as in determining the efficiency of antioxidant additives. The method used for the determination of oxidative stability is known as Rancimat and involves exposing the sample to an airflow at a
3
high temperature, usually 110 ºC. Under these conditions, the formation of the resulting compounds from oxidation is intensified, and the volatiles are conducted to measure cells containing deionized water, whose conductivity is continuously measured. Oxidative stability is expressed by the induction period (in hours), being determined at the point when the conductivity level, in the collecting flask, increases rapidly, signaling the onset of the overall oxidation process [14]. Water content may promote oil or biodiesel degradation and the proliferation of microorganisms, affecting its quality [15]. Karl Fischer titration is the most frequently used method for measuring the water content (humidity) in oils, solvents and other products. The method involves the titration with Karl Fischer reagent of a sample dissolved in methanol [16]. In this context, considering the large number of analyses, as well as the disadvantages of the use of toxic solvents and waste generation, multivariate calibration is suggested to reduce the number of tests. Therefore, in this work, hydrogen nuclear magnetic resonance (1H NMR) spectra were used to build multivariate models, from partial least squares (PLS), in order to perform simultaneous determination of six important quality parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability and water content. From 1H NMR, it is possible to extract a lot of informations about organic compounds. Despite the considerable cost of a NMR spectrometer, there are many advantages to using this technique. For example: qualitative/quantitative analysis can be done simultaneously; measurements can be performed quickly (for an abundant nucleus, such as 1H); it is not required to isolate the analyte when present in mixtures; analysis can be performed simultaneously for different analytes in a
4
single spectrum [17]. Furthermore, there is often NMR equipment for multiple users available in many institutions.” 2. Materials and methods 2.1. Biodiesel Samples Biodiesels from twelve different vegetable oil sources were used, such as babassu, brown flaxseed, canola, corn, cottonseed, macauba kernel, microalgae, palm, residual frying, sesame, soybean, and sunflower. Refined oils were employed for biodiesel synthesis without further purification. For the unrefined oils and the residual frying oil, because of their high acid value, it was necessary to perform acid esterification before transesterification in order to decrease the quantity of free fatty acids. The transesterification reaction was performed using the optimized procedure as previously reported, using analytical grade methanol as the alcohol and sodium methoxide, 30% w/w in methanol, as the catalyst [18]. 2.2. 1H NMR Spectra All of the 1H NMR spectra were recorded in a Bruker Avance III 500 MHz NMR spectrometer. To obtain the spectra, a sample of 20 µL of each biodiesel was dissolved in 600 µL of CDCl3, containing tetramethylsilane (TMS) as an internal reference (0.00 ppm), using the following experimental conditions: pulse program, zg30; spectral width, from –4.00 to 16.00 ppm; spectral size, 32768 points; 90º pulse, 11.75 µs; delay, 5 s; and number of scans, 16. The spectra were obtained in triplicate for each sample, and only 2 minutes was taken to obtain each spectrum. 2.3. Density at 20 ºC
5
Slightly less than 5 mL of the sample was added to a calibrated 5 mL volumetric flask of known weight. The flask with the sample was thermostated in a water bath at 20.0 ± 0.1 ºC for 10 minutes. Then, the flask was brought to volume at this temperature. The flask was dried with paper towels and weighed. The density was calculated based on the mass/volume relationship. 2.4. Kinematic Viscosity at 40 ºC by ASTM D445 The kinematic viscosity was determined according to ASTM D445 [19], using glass kinematic viscometer tubes Cannon Fenske and a thermostated bath Quimis. 2.5. Iodine Value by Wijs Method (BS EN 14111) The iodine value was determined according to BS EN 14111, using a 50 mL digital manual burette [20]. 2.6. Acid Number by AOCS Cd 3d-63 The acid number was determined according to AOCS Cd 3d-63 [21]. To perform the titration, a Metrohm Titrando 809 was used with a Metrohm Solvotrode electrode (electrolytesaturated LiCl in ethanol) and an automatic sampler Metrohm 814 USB Sample Processor. 2.7. Oxidative Stability at 110 ºC and under 10 L h-1 air flow by Rancimat (BS EN 14112) The induction period, which expresses the oxidative stability, was determined through the BS EN 14112 method using a Metrohm Rancimat model 873 [22].
2.8. Water Content by Coulometric Karl Fischer Titration (BS EN ISO 12937)
6
The total water content in biodiesel was determined according to BS EN ISO 12937 [23], using Karl Fischer coulometric titration (Metrohm model 831). 2.9. Chemometric Procedures by Partial Least Squares (PLS) PLS was performed using the Pirouette® 4.5 software. First, as pretreatment, each 1
H NMR spectrum was normalized from 0 to 1, dividing all points of each spectrum by its
highest value, to mitigate the influence of the biodiesel concentration in the solvent (CDCl3). Then, the first derivative of each spectrum was obtained to avoid influences of baseline variations. Finally, as preprocessing, the spectra were mean centered. This procedure involves the subtraction of the element of each column (variable) by the average value of the elements of that column (variable), the result is a matrix in which all columns (variables) have zero mean. Besides this preprocessing, the variables related to the solvent and TMS were excluded to perform the procedures since these spectral regions do not bring information related to the samples. Therefore, it was employed spectral width from 0.50 to 5.70 ppm (16908 points).
3. Results and discussion Table 1 shows the results obtained for the six quality parameters of the biodiesel samples. The different types of esters present in biodiesel, which generate different quality parameters, could be differentiated by the 1H NMR spectra. These differences are noted mainly by the number of unsaturations, the positions, and the length of the ester carbon chains. [Table 1]
7
The 1H NMR spectra obtained for all biodiesel samples are shown overlaid in Fig. 1 to illustrate the type of information that was used to build the multivariate models, using the PLS chemometric tool. [Fig. 1] The PLS regression method is a multivariate calibration procedure that uses Principal Component Analysis (PCA) to reduce the size of the data set and a correlation is made between this data set, such as spectra, and properties of interest, wherein these properties of interest can be physical or chemical characteristics. As in the calculation of principal components (PC), the property values of interest are taken into account (supervised method), the PC are named latent variables (LV) [24]. Figure 2 shows the graphs for reference values versus predicted values by calibration and by cross-validation. [Fig. 2] In Fig. 2, appropriate calibration can be noted, with correlation coefficients of calibration (rcal) being higher than 0.99, and suitable cross-validation with correlation coefficients of crossvalidation (rcv) being higher than 0.98. This indicates accentuated correlations between the 1
H NMR spectra and all of the studied parameters. In Table 2, the external validation results for the six quality parameters are presented as
the average of three replicates. Validations were performed considering each type of biodiesel, not each replicate. In order to calculate the relative errors, the mean measured values were considered as the true values.
8
[Table 2] The highest mean relative error value in the external validation was 12% for acid number; however, this parameter showed very low values. For instance, the maximum absolute error between the reference value and the correspondent predicted value is about 0.04 mg KOH g-1 that indeed is quite low for this attribute. In the other cases, the mean relative errors are low enough to allow the use in the determination of the related proprieties. Therefore, the PLS models proposed above can be recommended for the screening analysis of biodiesel concerning six quality parameters by linking 1H NMR spectra and chemometrics. In general, screening analysis is related to methods that normally involve little or no sample treatment, with the fast acquisition of qualitative, semi-quantitative, or even quantitative data about many characteristics of the sample; their response is used for decision-making, and oftentimes, confirmation by conventional methods is unnecessary [25]. The multivariate model accuracy is expressed by root mean square error (RMSE) of calibration (RMSEC), of cross-validation (RMSECV), obtained by internal validation, and of prediction (RMSEP), which is calculated from an external validation set of samples. Equation 1 represents the general equation of errors: n
(y yˆ ) i
RMSE
2
i
i1
n
Where: n = the number of samples; ̂ = predicted value for i-th sample;
(1) = reference
value for i-th sample. All of these kinds of errors are presented in Table 3, where the correlation coefficients of calibration (rcal), validation (rcv), number of latent variables (LV), and working range for all
9
proposed multivariate models are shown. The minimum number of LV was chosen based on RMSECV values, calculated for each LV. As shown, for example, in Figure 3, RMSECV values are almost constant from eighth LV for the PLS model that predicts acid number. For the other models, the number of LV was chosen analogously. [Table 3] [Figure 3] The parameters in Table 3 confirm the suitable calibrations, internal validations and predictive power of the chemometric models. As a result, as well as determining the molecular structures of organic compounds, the 1H NMR spectrum can provide information about several quality parameters of its compounds when chemometric tools are applied. From the results obtained (Table 2 and 3), it is possible to observe a better performance of the models to determine viscosity, density, and iodine value, because they showed suitable mean relative differences (Table 2) and it was not necessary to exclude samples to build the models, so that the entire working range among the samples was used. These parameters are more directly related to the major compounds present in biodiesel (methyl esters); because of this, they were probably easier to calibrate, validate, and predict. On the other hand, to build the models for acid number, induction period, and water content, some samples must be excluded to get an appropriate adjustment, because they presented much higher or much lower parameters compared with the other samples. In addition, the induction period is a parameter that depends not only on the structure of the methyl esters. Impurities from the synthesis, for example, alcohol, or the presence of metal ions or other organic compounds may influence the oxidative stability, even at low concentrations [26]. Fatty
10
acids can be detected and differentiated from the methyl esters by 1H NMR spectra [17]; however, they are present in biodiesel at very low concentrations. In this context, it is believed that the models to determine induction period and acid number showed higher mean relative errors for these reasons (Table 2). From these results, the application of the developed PLS models is proposed for the screening analysis of biodiesel parameters. These models allow the simultaneous analysis of six parameters, reducing analysis time, reagent consumption and therefore waste generation. Some multivariate models have been built for the determination of biodiesel quality parameters, such as density, water content, sulfur content, oxidative stability, and residual alcohol, among others [27-31]. However, all of them use infrared spectroscopy as the source of data. In this context, the development of other kinds of multivariate models, from 1H NMR spectroscopy, is interesting as they offer more analytical possibilities.
4. Conclusion This study demonstrated correlations between 1H NMR data with six studied parameters of biodiesel: density at 20 ºC, kinematic viscosity at 40 ºC, iodine value, acid number, oxidative stability and water content. Therefore, a screening analysis to enable the simultaneous determination of these six quality parameters has been developed from 1H NMR and PLS. The method showed good correlation, accuracy, and low relative errors. Once the multivariate model is built and implemented, it determines at least six important parameters of biodiesels, using simple and rapid procedures. Also, it minimizes experimental steps, analysis time, the consumption of reagents and solvents and waste generation.
11
Acknowledgements The authors are grateful to the Brazilian National Council of Technological and Scientific Development (CNPq) for financial support.
References [1] A.E. Atabani, A.S. Silitonga, I.A. Badruddin, T.M.I. Mahlia, H.H. Masjuki, A.S. Mekhilef, A comprehensive review on biodiesel as an alternative energy resource and its characteristics, Renewable Sustainable Energy Rev. 16 (2012) 2070-2093. [2] D.Y.C. Leung, B.C.P. Koo, Y. Guo, Degradation of biodiesel under different storage conditions, Bioresour. Technol. 97 (2006) 250-256. [3] I.P. Lôbo, S.L.C. Ferreira, R.S. Cruz, Biodiesel: parâmetros de qualidade e métodos analíticos, Quim. Nova 32 (2009) 1596-1608. [4] ASTM D6751 – 15a, Standard specification for biodiesel fuel blend stock (B100) for middle distillate fuels, 2015. [5] BS EN 14214:2012+A1:2014, Liquid petroleum products – fatty acid methyl esters (FAME) for use in diesel engines and heating applications – requirements and test methods, 2014. [6] Agência Nacional do Petróleo, Gás Natural e Biocombustíveis (ANP), Resolução Nº 45, 2014.
12
[7] B. Tesfa, R. Mishra, F. Gu, N. Powles, Prediction models for density and viscosity of biodiesel and their effects on fuel supply system in CI engines, Renewable Energy 35 (2010) 2752-2760. [8] G. Knothe, K.R. Steidley, Kinematic viscosity of biodiesel fuel components and related compounds. Influence of compound structure and comparison to petrodiesel fuel components, Fuel 84 (2005) 1059-1065. [9] G.G. Shimamoto, M.M.A. Favaro, M. Tubino, Simple methods via mid-IR or 1H NMR spectroscopy for the determination of the iodine value of vegetable oils, J. Braz. Chem. Soc. 26 (2015) 1431-1437. [10] G. Knothe, Structure indices in FA chemistry. How relevant is the iodine value?, J. Am. Oil Chem. Soc. 79 (2002) 847-854. [11] G.G. Shimamoto, J.A. Aricetti, M. Tubino, A simple, fast, and green titrimetric method for the determination of the iodine value of vegetable oils without Wijs solution (ICl), Food Anal. Methods 9 (2016) 2479–2483. [12] M. Tubino, J.A. Aricetti, A green potentiometric method for the determination of the iodine number of biodiesel, Fuel 103 (2013) 1158-1163. [13] M. Tubino, J.A. Aricetti, A visual titration method for the determination of the acid number of oils and fats: A green alternative, J. Am. Oil Chem. Soc. 89 (2012) 2113-2115. [14] J. Pullen, K. Saeed, An overview of biodiesel oxidation stability, Renewable Sustainable Energy Rev. 16 (2012) 5924-5950.
13
[15] G. Knothe, Some aspects of biodiesel oxidative stability, Fuel Process. Technol. 88 (2007) 669-677. [16] A. Felgner, R. Schlink, P. Kirschenbühler, B. Faas, H.D. Isengard, Automated Karl Fischer titration for liquid samples – Water determination in edible oils, Food Chem. 106 (2008) 13791384. [17] J.H. Simpson, Organic structure determination: using 2-D NMR spectroscopy: a problembased approach, Elsevier/Academic, Amsterdam, 2008. [18] W.L.G Silva, P.T. Souza, G.G. Shimamoto, M. Tubino, Separation of the glycerol-biodiesel phases in an ethyl transesterification synthetic route using water, J. Braz. Chem. Soc. 26 (2015) 1745-1750. [19] ASTM D445–12, Standard test method for kinematic viscosity of transparent and opaque liquids (and calculation of dynamic viscosity), 2012. [20] BS EN 14111, Fat and oil derivatives – fatty acid methyl esters (FAME) – determination of iodine value, 2003. [21] American Oil Chemists' Society, AOCS Official Method Cd 8d-63, in: D. Firestone (Ed), Official methods and recommended practices of the AOCS, AOCS, Champaign, 1973. [22] BS EN 14112, Fat and oil derivatives. Fatty acid methyl esters (FAME). Determination of oxidation stability (accelerated oxidation test), 2003. [23] BS EN ISO 12937, Methods of test for petroleum and its products. BS 2000-438. Petroleum products. Determination of content. Coulometric Karl Fischer titration method, 2001.
14
[24] K.R. Beebe, R.J. Pell, M.B. Seasholtz, Chemometrics: a practical guide, John Wiley & Sons, New York, 1998. [25] R. Muñoz-Olivas, Screening analysis: an overview of methods applied to environmental, clinical and food analyses, Trends Anal. Chem. 23 (2004) 203-216. [26] N. Canha, P. Felizardo, J.C. Menezes, J.N. Correia, Multivariate near infrared spectroscopy models for predicting the oxidativestability of biodiesel: Effect of antioxidants addition, Fuel 97 (2012) 352-357. [27] M. Bampi, A.P. Scheer, F. Castilhos, Application of near infrared spectroscopy to predict the average droplet size and water content in biodiesel emulsions, Fuel 113 (2013) 546-552. [28] P. Felizardo, P. Baptista, J.C. Menezes, M.J.N. Correia, Multivariate near infrared spectroscopy models for predicting methanol and water content in biodiesel, Anal. Chim. Acta 595 (2007) 107-113. [29] R.M. Balabin, R.Z. Safieva, Near-infrared (NIR) spectroscopy for biodiesel analysis: fractional composition, iodine value, and cold filter plugging point from one vibrational spectrum, Energy Fuels 25 (2011) 2373-2382. [30] L.F.B. Lira, M.S. Albuquerque, J.G.A. Pacheco, T.M. Fonseca, E.H.S. Cavalcanti, L. Stragevitch, M.F. Pimentel, Infrared spectroscopy and multivariate calibration to monitor stability quality parameters of biodiesel, Microchem. J. 96 (2010) 126-131.
15
[31] L.F.B. Lira, F.V.C. Vasconcelos, C.F. Pereira, A.P.S. Paim, L. Stragevitch, M.F. Pimentel, Prediction of properties of diesel/biodiesel blends by infrared spectroscopy and multivariate calibration, Fuel 89 (2010) 405-409.
16
Figure Captions Fig. 1. Overlapping 1H NMR spectra of biodiesel samples.
Fig. 2. Relationship between the reference and the predicted values obtained by calibration and by cross-validation via PLS model for density, kinematic viscosity, iodine value, acid number, induction period and water content of biodiesel samples.
Fig. 3. Root mean square error of cross-validation (RMSECV) as function of the number of latent variables (LV) for the PLS model that predicts acid number.
17
Tables Table 1. Quality parameters of biodiesel samples (average of three replicates). Kinematic Viscosity at 40 ºC
Density at 20 ºC
Biodiesel
/ (kg m-3)
/ (mm2 s-1)
Iodine Value
Acid Number
/ (g I2 100 g-1)
/ (mg KOH g-1)
Induction Period
Water Content
/ (h)
/ (mg kg-1)
Babassu
868.13 0.07
2.713 0.006
15.2 0.4
0.670 0.008
10.2 0.3
178 6
Brown flaxseed
884.36 0.03
3.61 0.04
151 6
0.117 0.003
1.91 0.03
193 6
878.0 0.2
5.43 0.04
106.2 0.6
0.22 0.02
9.2 0.3
270 20
Corn
879.08 0.01
4.45 0.03
114.6 0.4
0.188 0.008
7.10 0.03
277 7
Cottonseed
879.25 0.06
4.510 0.009
109 1
0.16 0.01
3.06 0.06
620 30
869.4 0.2
3.003 0.004
30 2
0.204 0.008
61 6
190 20
Microalgae
878.20 0.03
4.660 0.006
107 1
0.15 0.01
7.62 0.09
198 8
Palm
873.18 0.03
3.44 0.02
20.4 0.4
0.240 0.003
55 3
394 5
Residual frying
888.0 0.2
4.616 0.007
113.5 0.5
0.17 0.01
1.61 0.02
390 10
Sesame
879.1 0.1
4.240 0.001
104.9 0.2
0.11 0.01
8.7 0.1
80 20
Soybean
883.6 0.1
4.38 0.01
124.3 0.9
0.201 0.008
6.2 0.4
249 4
Sunflower
885.9 0.1
4.718 0.008
123.4 0.5
0.222 0.002
2.1 0.1
233 6
Canola
Macauba kernel
Table 2. Predicted and measured parameters values for external samples (averages of three replicates) and relative errors. Density at 20 ºC / (kg m-3) Biodiesel
Kinematic Viscosity at 40 ºC / (mm2 s-1)
Iodine Value / (g I2 100 g-1)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Corn
881.55
879.08
0.3
4.63
4.45
4.0
114.9
114.6
0.3
Soybean
887.3
883.6
0.4
4.35
4.38
0.7
121.2
124.3
2.5
Sunflower
884.4
885.9
0.2
4.378
4.718
7.2
127.3
123.4
3.1
Mean Relative Error (%)
Mean
0.3
Mean
Acid Number / (mg KOH g-1) Biodiesel
4.0
Mean
Water Content / (mg kg-1)
2.0
Induction Period / (h)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Mean Predicted Value
Mean Measured Value
Relative Error (%)
Corn
0.196
0.188
4.3
276
277
0.4
7.3
7.1
2.8
Soybean
0.181
0.222
18
255
249
2.4
5.9
6.2
4.8
Sunflower
0.171
0.201
15
237
233
1.7
2.36
2.14
11
Mean Relative Error (%)
Mean
12
Mean
1.5
Mean
6.2
18
Table 3. Parameters of PLS models to determine density, kinematic viscosity, iodine value, acid number, induction period and water content. Partial Least Squares (PLS) Models Iodine Value
Acid Number a
Induction Period b
Water Content c
868.13 – 887.97 (kg m-3)
Kinematic Viscosity at 40ºC 2.713 – 5.43 (mm2 s-1)
15.2 – 151 (g I2 100 g-1)
0.11 – 0.240 (mg KOH g-1)
1.61 – 9.2 (h)
178 – 394 (mg kg-1)
12
12
12
11
9
10
36
36
36
33
27
30
16908
16908
16908
16908
16908
16908
Parameter Density at 20ºC Working Range Types of Biodiesel (n=3 replicates) Samples Number of Variables (from 0.5 to 5.7 ppm) Latent Variable
8
7
-3
RMSEC
0.4 (kg m )
rcal
0.9984
RMSECV
0.6 (kg m-3)
rcv
0.9955
RMSEP
3 (kg m-3)
4 8 6 8 3 0.003 0.08 (mm s ) 0.1 (h) 6 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) 0.9958 0.9985 0.9978 0.9994 0.9974 3 0.004 2 -1 0.09 (mm s ) 0.6 (h) 12 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) 0.9930 0.9976 0.9956 0.9835 0.9871 3 0.03 2 -1 0.2 (mm s ) 0.2 (h) 8 (mg kg-1) (g I2 100 g-1) (mg KOH g-1) excluded to build the model, because its acid number is much higher compared to 2 -1
a = babassu biodiesel sample was others. b = babassu, macauba, and palm biodiesel samples were excluded to build the model, because their induction period values are much higher compared to others. c = cottonseed and sesame biodiesel samples were excluded to build the model, because their water content values are much different compared to others.
19
20
Highlights
1
The models can perform simultaneous quantitative analysis of six biodiesel
H NMR spectra were used to build multivariate models
parameters
Parameters: density, viscosity, iodine value, acid number, oxidative stability, water.
There is significant reduction of analysis time, reagents consumption and waste.
21