Using NIR Spectroscopy for Real-Time Inline Predictions of Jet Fuel Properties

A variety of chemical and physical properties of jet fuels are used to characterize their grade and quality. These properties affect the use and performance of these fuels. These properties are quantified and reported by costly and elaborate experimental techniques that take significant amount of time. These methods are often based on standardized techniques specified by the ISO and ASTM. Variations in the characteristic physical properties of jet fuel (viscosity, freezing point, etc.) are known to change based on the variations in the fuel’s chemical composition. This link between the chemical composition and physical properties helps in making precise predictions of these physical properties by correlating with results acquired from direct chemical analysis.

Near Infrared (NIR) spectroscopy is a chemical analysis technique that employs NIR light to record a picture of a wide range of chemical properties, namely the presence of different kinds of chemical bonds or chemical components. The chemical measurements recorded by NIR spectroscopy can then be correlated to the defining physical properties of jet fuel. In addition, scattering effects recorded during an NIR experiment correlate directly to physical properties.

Sample preparation is not required for a non-destructive NIR measurement. This measurement is performed by studying liquid or solid samples as-is. Reagents, solvents, and heating or cooling are not required. This analysis can be performed using a benchtop instrument in a laboratory, or alternatively fiber optics can be used in-process to acquire real-time measurements. In addition, a single spectrum can be used to predict multiple properties.

In this analysis, NIR techniques were developed and validated for Aromatics (V%), API Gravity, Density (15°C), Cetane Index, flash point (°C), boiling profiles such as temperatures at 10%, 20%, 50%, and 90% distillation recovery, freezing point (°C), viscosity at - 20°C (CSt), and hydrogen content (% weight). Metrohm NIRS XDS process analyzer fitted with an immersion probe was used to calibrate and validate the samples. These parameters demonstrated excellent correlation between the spectral difference and the parameter changes. For individual parameter, correlation coefficients, standard error of prediction (SEP), standard error of cross validation (SECV), and standard error of calibration (SEC) are reported for each parameter. These results demonstrate that NIR spectroscopy can be utilized to make accurate predictions of all these parameters from a single measurement that takes as less as 30 seconds. This precision corresponds with the accuracy of the ASTM technique defined for all the parameters. The data obtained in real time can be used to track and obtain a reliable product quality.

Samples

109 samples were studied. The samples include a variety of jet fuel grades (Jet A, Jet A-1, JP5, JP8, etc.). NIR and the elated ASTM technique were used for samples analysis. While all the 12 parameters were unavailable for all samples, each parameter exhibited a minimum of 62 samples for model development. Table 1 shows the minimum and maximum values, the number of samples, and ASTM reproducibility for individual parameters.

Table 1. Number of samples, min value, max value, and ASTM precision, NIR standard error of calibration (SEC), NIR standard error of cross-validation (SECV), and NIR standard error of prediction (SEP) for each parameter correlated to NIR data

Samples Max Min ASTM Precision NIR SEC NIR SECV NIR SEP R2
API Gravity 104 48.2 37.8 0.3 0.26 0.27 0.29 0.986
Density @15°C (kg/L) 104 0.8353 0.7874 0.0012 0.0013 0.0013 0.0016 0.984
Aromatics V% 99 24.4 10.8 2.63 0.50 0.51 0.72 0.962
CetaneIndex 101 49.0 33.0 <2.0 0.85 0.89 0.84 0.934
°[email protected]% Rec 104 210.3 153.9 3.6 3.2 3.3 3.6 0.879
°[email protected]% Rec 73 203.9 166.0 NA 1.9 1.9 2.0 0.952
°[email protected]% Rec 104 228.5 185.0 2.97 2.2 2.2 2.4 0.927
°[email protected]% Rec 104 273.5 157.8 3.6 3.6 3.6 3.6 0.839
Flash Point °C 105 78.3 38.0 4.3 1.9 2.1 2.3 0.925
Freeze Point °C 104 -40.6 -65.5 0.8 1.9 2.1 2.0 NA
Hyd Content wt% 98 14.20 13.28 0.16 0.05 0.05 0.05 0.939
Vis @ 20 °C (cSt) 95 12.440 1.700 0.0694 0.2139 0.2172 0.2331 0.905

Instruments Used

  • Microbundle 4 Channel
  • NIRS XDS process analyzer
  • Probes and accessories - Immersion Probe SS, NIRS Microbundle Interactance
  • NIR data collection parameters - Reference Std (standardized); collection range (800-2200nm); scans per spectrum (32/32)

Sample Analysis

Samples were placed in a sample container. The Metrohm NIRS XDS process analyzer equiped with a stainless steel immersion probe was used for analysis (Figure 1). The NIR probe was immersed into the sample and a spectrum was obtained by transflectance. For each individual sample, a total of three spectra were collected after stirring the sample before each acquisition.

Metrohm NIRS XDS Process Analyzer -Microbundle equipped with stainless steel immersion probe for liquid analysis.

Figure 1. Metrohm NIRS XDS Process Analyzer -Microbundle equipped with stainless steel immersion probe for liquid analysis.

Quantitative Method

A quantitative model was developed using the Vision® chemometric software with the spectra and the related ASTM values. It is essential to perform the 2nd derivative pre-math treatment to create an optimum correlation between the ASTM values and the spectral difference. In order to optimize the partial least squares regression, a leave-one-out cross validation was employed. The method was the tested using an independent validation set.

Results and Discussions

Figure 2 shows the overlay of NIR spectra of the jet fuel samples. The absorption bands match with vibrational combinations and overtones of varied types of chemical bonds. SAll the samples shared similar absorption features because overall the chemical makeup of each samples is similar.

Raw NIR spectra of jet fuel samples acquired with the Metrohm NIRS XDS Process Analyzer equipped with a stainless steel immersion probe.

Figure 2. Raw NIR spectra of jet fuel samples acquired with the Metrohm NIRS XDS Process Analyzer equipped with a stainless steel immersion probe.

However. small chemical variations that affect the jet fuel‘s physical properties create slight variations in the NIR spectra. Such changes can be associated with the differences in physical properties by utilizing the method illustrated in the quantitative method section. As soon as consistent correlations have been determined, a range of physical properties of analogous samples can be predicted based on calibration equations and the NIR spectrum.

For each parameter, PLS regression techniques were developed and a calibration technique was developed and tested on a validation set for individual parameters. Figures 3 to 15 show that the NIR predictions made for each parameter are compared to the ASTM values for each sample.

Second Derivative Spectra of jet fuel samples acquired with the Metrohm NIRS XDS Process Analyzer equipped with a stainless steel immersion probe.

Figure 3. Second Derivative Spectra of jet fuel samples acquired with the Metrohm NIRS XDS Process Analyzer equipped with a stainless steel immersion probe.

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for API Gravity calibration set (left) and validation set (right).

Figure 4. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for API Gravity calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Density calibration set (left) and validation set (right).

Figure 5. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Density calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Aromatics (% volume) calibration set (left) and validation set (right).

Figure 6. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Aromatics (% volume) calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Cetane Index calibration set (left) and validation set (right).

Figure 7. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Cetane Index calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D10% calibration set (left) and validation set (right).

Figure 8. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D10% calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D20% calibration set (left) and validation set (right).

Figure 9. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D20% calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D50% calibration set (left) and validation set (right).

Figure 10. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D50% calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D90% calibration set (left) and validation set (right).

Figure 11. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for D90% calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Flash Point (°C) calibration set (left) and validation set (right).

Figure 12. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Flash Point (°C) calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for FreezingPoint (°C) calibration set (left) and validation set (right).

Figure 13. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for FreezingPoint (°C) calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Hydrogen Content (% wt.) calibration set (left) and validation set (right).

Figure 14. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Hydrogen Content (% wt.) calibration set (left) and validation set (right).

NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Viscosity @ -20 °C (cSt) calibration set (left) and validation set (right).

Figure 15. NIR Predictions (y-axis) compared to ASTM laboratory values (x-axis) for Viscosity @ -20 °C (cSt) calibration set (left) and validation set (right).

The SEP, SEC and SECV are compared against the ASTM precision in Table 1 for individual parameter. Also listed is the correlation coefficient (R2). The following sections discuss the results for each parameter.

The standard method of analysis (ASTM D1298 and ASTM D287) uses a hydrometer and a series of calculations and corrections to establish the American Petroleum Institute (API) gravity of jet fuel at 60°F. API gravity is a relative parameter associated with density. In this analysis, ASTM D1298 was used. The dimensionless values correspond with water density, but at times are reported as degrees. Values less than 10 suggest that the substance is heavier than water and values greater than 10 are lighter than water. API gravity for jet fuel is often between 37 and 51. The NIR technique for API led to SEP of 0.29 that corresponds with the accuracy of the ASTM technique. The absolute density can be reported instead of a relative value. Using the API gravity (ASTM D1298) or a digital density meter (ASTM D4052), absolute density can be measured. Here, a digital density meter was utilized. The density is usually reported at 15°C. The jet fuel’s density spans between 0.775 and 0.840kg/L, and the precision of the NIR predictions less than this range was 0.0016kg/L, which relates well with the accuracy of the digital density meter method (0.0012 kg/L).

While aromatics have high energy in fuel, they tend to create soot when burned. The quantity of aromatics in jet fuel can differ considerably based on how the jet fuel was derived. Aromatics less than 20 to 25% ensure optimum performance. The existing method used for determining aromatics is time consuming and involved column separation and detection method known as fluorescent indicator adsorption analysis (ASTM D1319). The rapid NIR predictions gave a accuracyof 0.72% by volume. This is suspiciously lower than the precision of the ASTM method (2.63%).

There is no minimum cetane index in jet fuel as this index is relevant to compression ignition engines, but it is still possible to determine and report this index as this forms a critical parameter for petroleum products. Costly combustion chambers (ASTM D613) are used to determine the cetane index, but the value can be measured using mid-boiling point (D976) and API gravity. The quantified cetane index can span from 30- 60, and the minimum for diesel products is about 40. The NIR accuracy of 0.84 agrees well with the predicted accuracy of the ASTM method of less than 2.0.

The longest used test technique for petroleum product is the determination of the boiling range using a simple atmospheric batch distillation. The distillation of jet fuel is a critical property that impacts the safety and performance of the product. In the test procedure defined by ASTM D86, n% by volume of heated sample is recovered from a stage distillation. Jet A and Jet A-1 fuel have the highest distillation temperature of 205°C at 10% recovery, while other grades require lower temperatures. Temperatures at 10%, 20%, 50%, and 90% distillation recovery were specified for the samples used here. NIR predicted the DX% temperatures with a accuracy corresponding to the ASTM method.

A key defining property of jet fuel is the flash point, which is used for classifying different grades of jet fuel. The lowest temperature is the flash point, where the jet fuel would vaporize and burn in the presence of oxygen. Based on the grade, jet fuel may need flash points greater than 60°C for certain military grades and higher than 38°C for jet A and B. Based on the method used, results for the different standard test methods may differ to a large extent. The ASTM D93 method was used for the study. The precision of the NIR prediction (2.3°C) was lower than the predicted accuracy of the ASTM method (4.3°C).

The freezing point refers to the temperature at which stage crystals are formed in the fuel. In addition to the flash point, the freezing point is a major parameter that is used to define the various grades of jet fuel. The ASTM D5972 and ASTM D2386 are standard methods used for establishing the jet fuel’s freezing point. Here, ASTM D5972 was used to determine the lab values reported. This was one of two parameters where the prediction accuracy of the NIR (2.0°C) was over 2 times larger when compared to the ASTM accuracy (0.8°C).

Specified as weight %, hydrogen content is directly related to combustion quality. Specifications for jet fuel need a minimum of 13.4 %H by weight. Typical techniques comprise experimental determination by nuclear magnetic resonance (ASTM D3701) or burning (ASTM D1018). In case these methods are unavailable, other measured parameters (ASTM D3343) can be used to predict the content. It was seen that the NIR predictions of hydrogen content has 0.05% precision which was relatively lower than the accuracy of the ASTM method (0.16%).

Viscosity can be thought of as the thickness of the jet fuel. A fuel that is highly viscous will exhibit varied flow properties when compared to a fuel of low viscosity. Hence, these fuels would have different performance in burner nozzles and pumps. Viscosity is often reported in centistokes (CSt). The ASTM D445 method determines the time for a volume of liquid to pass via a capillary viscometer. The laboratory values obtained for the samples used in this analysis span between 3.5 and 6.9 CSt. Using NIR, the viscosity can be estimated with a precision of 0.2331CSt which is fairly high in comparison to the accuracy of the original ASTM method (0.0694CSt).

Conclusion

In order to develop a good NIR model, care should be taken to ensure that the SEP is within 1.5 to 2.0X the error of the lab/primary method (SEL). As the error of the laboratory technique is inherent in the NIR calibration, the SEP should not be lower than the SEL. In this analysis, it was seen that the SEP for NIR was lower than that of the SEL with respect to the flash point, aromatics, and hydrogen content. In such instances, it may appear that the precision of the primary technique was higher than usual as a result of careful processes. In contrast, the high SEP for viscosity and freezing point could be the result of high sensitivity to outlier sample and other NIR factors.

This information has been sourced, reviewed and adapted from materials provided by Metrohm AG.

For more information on this source, please visit Metrohm AG.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Metrohm AG. (2019, April 02). Using NIR Spectroscopy for Real-Time Inline Predictions of Jet Fuel Properties. AZoM. Retrieved on August 21, 2019 from https://www.azom.com/article.aspx?ArticleID=12539.

  • MLA

    Metrohm AG. "Using NIR Spectroscopy for Real-Time Inline Predictions of Jet Fuel Properties". AZoM. 21 August 2019. <https://www.azom.com/article.aspx?ArticleID=12539>.

  • Chicago

    Metrohm AG. "Using NIR Spectroscopy for Real-Time Inline Predictions of Jet Fuel Properties". AZoM. https://www.azom.com/article.aspx?ArticleID=12539. (accessed August 21, 2019).

  • Harvard

    Metrohm AG. 2019. Using NIR Spectroscopy for Real-Time Inline Predictions of Jet Fuel Properties. AZoM, viewed 21 August 2019, https://www.azom.com/article.aspx?ArticleID=12539.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this article?

Leave your feedback
Submit