The capability to quantify trace contaminants in biodiesel is crucial for optimizing the biodiesel production process and ensuring final product quality. The base-catalyzed transesterification reaction of an oil or fat is the most commonly used process for making biodiesel. The oil (triglyceride) is reacted with excess alcohol in the presence of a catalyst (potassium hydroxide or sodium hydroxide) to produce biodiesel and glycerin. The transesterification reaction is efficient with yields approaching 100 percent. If the reaction worked perfectly, only biodiesel and glycerin would be present after the reaction. Any trace contaminants such as glycerin, glycerides, water, methanol or free fatty acids in the biodiesel indicates that some step in the process, including the reaction, is not optimized.

Contaminants that are of most importance for biodiesel final product quality are free and total glycerin, water and methanol since final biodiesel product must meet ASTM D 6751-07 specifications. The purpose of this article is to demonstrate that Fourier transform near-infrared technology can quickly and accurately quantify the concentration of contaminants in biodiesel.

Quantifying impurities in the biodiesel production process, either in-line or online, can help optimize the yield and purity of the fuel. There are multiple points in the process where near-infrared technology works well for predicting complex matrices in-line or online, an advantage over traditional lab methods. Often a single, centrally located instrument using fiber optic cables and probes can be used to analyze multiple key process points. This eliminates the need for the operator to take samples to a common laboratory for analysis.

Since the results are produced in real-time, the plant personnel can quickly make adjustments to the process, and often closed-loop control strategies can be used to automate process adjustments. The real-time process results can also be plotted to show trends, or be entered into databases for further statistical processing.

The ability to quickly analyze samples for multiple components using Fourier transform near-infrared results in time savings over traditional lab methods. In this study, a lab-based Fourier transform near-infrared analyzer was able to generate results in 30 seconds. Alternative methods for determining free and total glycerin can take up to one hour.

There are separate ASTM test methods under ASTM D 6751-07 for quantifying contaminants present in biodiesel. They all require different equipment, reagents and supplies, which increase the cost and add to the overall complexity of the quality assurance system. Methanol presence is done by a flash point tester or a gas chromatography method. The flashpoint method does not allow for exact quantification of how much methanol is present, only that the concentration is above or below a certain threshold. The amount of water present is determined by the percentage volume of water centrifuged out of a sample. Free fatty acid levels are determined by a potentiometric titration that is not part of the ASTM test methods.

The Experiment
The starting material for preparing standards was a pure biodiesel sample from a plant that uses soybean oil to make biodiesel. This pure sample along with three other samples saturated with glycerin and/or water were used to form four stock standards. These stock solutions were combined to produce lots that varied in their amount of biodiesel, water and glycerin. These lots had differing amounts of free fatty acids, methanol, tri-, di- and monoglycerides added to them to produce the standards used in the study. The component concentration in each standard was determined by percentage weight with each component addition having a mass uncertainty of plus or minus 0.0002. A total of 68 standards were prepared in order to cover a wide range of component concentrations. The values for glycerin and water were determined using accepted industry primary analytical methods. Pure standards of free fatty acids, methanol, tri-, di- and monoglycerides were purchased for standard preparation in this study.

For each type of glyceride (tri, di and mono), multiple pure glyceride standards were used for making the method standards. The glyceride standards differed based on the fatty acid chains that were attached to them. Most naturally occurring fatty acid chain lengths are 16, 18, or 20 carbons. In this study, a fatty acid chain length of 18 carbons was used. Table 1 shows the different pure glycerides and free fatty acids used in the standards.

The near-infrared spectral acquisition was performed in a laboratory using a trademarked Thermo Scientific Antaris II Fourier transform near-infrared analyzer. The spectral data was acquired in transmission using glass cuvettes and a heated cuvette holder set at 86 degrees Fahrenheit. A background was taken in between each sample scan.

The spectroscopic collection parameters were as follows:
› Spectroscopic range: 10,000 to 4,000 cm-1

› Resolution: 4 cm-1

› Co-averaged scans: 32 scans

› Collection time: 20 seconds

The calibration model was developed using the trademarked Thermo Scientific TQ Analyst software package for quantitative analysis. All spectra were converted into their second derivative prior to calibration. A partial least squares regression model was developed for this study because it gives accurate and robust methods for multicomponent mixtures.

The partial least squares model is a good choice for this application because it calibrates for each component separately, and it can effectively handle the high number of sources of variation due to the large number of standards with varying concentrations. TQ Analyst can be set to select spectral regions automatically by using concentration and spectral information of the calibration standards.

Data Analysis and Results
The fact that many of the compounds in this study have the same functional groups makes it hard to see spectral variation among the samples in the raw spectra. For this reason, a second derivative was taken since it increases the spectral peak definition while retaining peak location from the unprocessed spectra. The derivative also removes any baseline due to scattering. Since these samples were liquids run in transmission with a constant pathlength, scattering was not an issue.

Small changes in absorbance and spectral shape can be seen in the 4,900 to 5,000 cm-1 range which is the hydroxide combination band. The shift in peak and change in absorbance at 4,950 cm-1 is due to varying amounts of methanol in the standards. There is a distinct separation between the samples with the highest and lowest concentration of methanol.

The standards used to develop the calibration model had biodiesel concentration that varied from 90.1 percent to 99.9 percent. The contaminants made up the remaining weight percentage of the standard. As can be seen in Table 2, many trace contaminant concentrations in biodiesel can be predicted accurately. The root mean square error of calibration for all the components in the method was less than 0.2 percent and correlation coefficients were all greater than 0.93. The calibration plot for total glycerin, which is free glycerin plus the bound glycerin portion from the tri-, di- and monoglycerides, is the calculated (near-infrared model predicted values) versus the actual weight percentage for each standard. This plot shows good correlation from 0 percent to 1 percent. It also shows the independent validation points predict as well as the samples used for the calibration because they fall as close to the ideal prediction line (slope=1) as the calibration samples. The calibration plot for methanol shows good correlation from 0 percent to 2 percent and has independent validation points that are predicted well.

Conclusions
Fourier transform near-infrared can accurately predict concentrations of trace contaminants as well as fatty acid methyl ester in biodiesel as shown by this study. The ability to quickly quantify multiple trace contaminants allows for process adjustments to be made much more quickly than if samples were run by traditional primary analytical methods.

Transfering Fourier transform near-infrared methods from lab to line gives production facilities flexibility on instrument location and sample presentation. Transferring Fourier transform near-infrared inline allows process adjustments to be made in real-time using closed-loop control strategies. By optimizing the process, chemical and processing costs can be minimized. The capability to measure multiple sample points simultaneously with a multiplexing Fourier transform near-infrared makes inline measurement more practical and economical since one instrument might be all that is needed for a biodiesel production facility. The use of Fourier transform near-infrared simplifies the testing protocol for biodiesel process samples since one instrument can replace multiple pieces of equipment and eliminate lab supply costs. This study proves that Fourier transform near-infrared is a valuable tool for biodiesel process monitoring.

Stephanie Scherer and Warren Kosman are with Valparaiso University in Valparaiso, Ind. Chris Heil and Jeffrey Hirsch are with Thermo Fisher Scientific in Madison, Wis. For an extended version of this article, including additional charts and figures, visit http://www.BiodieselMagazine.com.