2486
Energy & Fuels 2006, 20, 2486-2488
Rapid Quantification of Kinematical Viscosity in Aviation Kerosene by Near-Infrared Spectroscopy Zhi-na Xing,* Ju-xiang Wang, Yong Ye, and Gang Shen Department of Aerocraft Engineering, NaVal Aeronautical Engineering Institute, Yantai 264001, China ReceiVed January 3, 2006. ReVised Manuscript ReceiVed August 6, 2006
A quantitative near-infrared spectroscopy (NIRS) method was developed for the determination of the kinematical viscosity of aviation kerosene in terms of chemometrics. Reference measurement was performed by ASTM445. The calibration model of aviation kerosene was established by partial least-squares (PLS) regression. Satisfactory calibration statistics were obtained with a standard error of calibration (SEC) of 0.012 mm2/s and a standard error of prediction (SEP) of 0.027 mm2/s. A t-test showed that the NIRS model had an accuracy equivalent to that of ASTM445 in the analysis of the kinematical viscosity of aviation kerosene. The study emphasized the potential of NIRS as a rapid and highly effective alternative analysis method to the conventional quantitative analysis of kinematical viscosity of aviation kerosene.
1. Introduction Kinematical viscosity is a significant parameter showing the thickness of aviation kerosene. The employed limit of this parameter is different according to different engines. If the kinematical viscosity oversteps, the fuel pump will work abnormally. While the traditional method (ASTM445) to determine the kinematical viscosity of aviation kerosene fulfills the necessary requirements of accuracy, specificity, and reproducibility, it is very time-consuming, requiring sophisticated equipment and a specific friable glass pipe. In short, it does not meet the need of a rapid, easy to use, selective, and sensitive analytical technique. As a result, a simple and rapid method, which is able to determine the kinematical viscosity of aviation kerosene accurately in real time, is necessary for the production and use of aviation kerosene. Near-infrared (NIR) absorption is attributed mainly to overtones and combinations of mid-IR vibrational bands involving N-H, O-H, and C-H bonds in molecules.1 The rich signal information from these groups makes near-infrared spectroscopy (NIRS) a useful technique for a lot of organic compounds.2-4 A few of opportune applications of NIRS in some fuels have been reported.5,6 Aviation kerosene comprises of a complex mixture of different types of organic compounds which can be determined by NIRS. And, the kinematical viscosity is strictly correlative to the components of aviation kerosene. Therefore, * Corresponding author. Address: 104 Staff Room, Naval Aeronautical Engineering Institute, Yantai, Shandong, China. Postcard: 264001. Tel.: 86+13371389388. E-mail:
[email protected]. (1) Skoog, D. A.; Holler, J. F.; Nieman, T. A. Principles of instrumental analysis, 5th ed.; Saunders College Publ: Philadelphia, 1998; pp 422430. (2) Williams, P. C.; Elharamein, F. J. Cereal Chem. 1986, 65 (2), 109112. (3) Renden, J. A.; Oates, S. S. Poultry Sci. 1986, 65 (8), 1539-1543. (4) Lu, W. Zh.; Yuan, H. F.; Xu, G. T. Modern NIRS analysis technology; petroleum publishing company of China: 2000; 4, pp 146-147. (5) Michael, D. J. The application of near-infrared spectroscopy for the quality control analysis of rocket propellant fuel premixes. Talanta 2004, 62, 675-679. (6) Vogelsanger, B.; Ossola, B. Proceedings of 31st International Annual Conference of ICT, Karlsruhe, Germany, 2000; pp 5-1-5-13.
NIRS should be a selective method to determine the kinematical viscosity of aviation kerosene. The aim of this paper was to describe the manner in which a robust NIRS method to determine kinematical viscosity was achieved and the results obtained with the new methodology. 2. Experimental Section 2.1. Instrumentation. NIR spectra were acquired using a NIR2000 spectra instrument (Yingxian instrument Co. Ltd., Beijing, China) with a transmission detector (CCD/2048 pixels) with a useable wavelength range of 700-1100 nm. All measurements were performed using a 5 cm path-length quartz cuvette at an invariable instrument temperature (37 ( 0.5 °C) and sample temperature (25 ( 0.5 °C) by two sets of electrical heated thermostats, respectively. Each NIR spectra was the average of 10 scans, using air as reference. Spectra were acquired over the 700-1100 nm range with the software provided with the instrument. Figure 1 shows the NIRS curve of aviation kerosene. 2.2. Sample Preparation. Here, 44 samples (1.54-1.78 mm2/s) were collected from three different manufactures of aviation kerosene in China. A group of 34 samples was taken to develop an NIR calibration model. The other 10 samples formed a validation set (viz., accuracy set) to test the calibration model. 2.3. Laboratory Reference Method. All samples were tested for kinematical viscosity with ASTM445 on a low temperature kinematical viscosity instrument (Model SYD-265G, Changji geology Co. Ltd., Shanghai, China). This was the laboratory reference method for analysis of samples employed as standards for NIRS calibration development.
3. Results and Discussion 3.1. Development of a Calibration Model. While the basic NIR absorption spectra may be used for calibration purposes, often the centering, 5 points smoothing, and second derivative of the sample spectra are used to remove the influence of random noise, decrease the systematic error, and increase the signalnoise ratio (SNR). Second derivatives were obtained using a segment size of 25 and a gap size of 0. The calibration algorithm for the calibration model was partial least-squares (PLS) which
10.1021/ef060003i CCC: $33.50 © 2006 American Chemical Society Published on Web 09/15/2006
Kinematical Viscosity in AViation Kerosene
Energy & Fuels, Vol. 20, No. 6, 2006 2487 Table 1. Parameters of the Calibration Model
modeling regression band (nm) 875-1065
spectra pretreatment 5 points smoothing, centering, second derivative (25)
was provided by the NIR system’s software package with the instrument. This algorithm involves the computation of linear combinations of various NIR wavelength contributions to produce factors. A certain minimum number of factors were
Figure 1. NIRS curve of aviation kerosene.
Figure 2. Effect of factor number on PRESS.
Figure 3. NIR prediction versus reference determination (n ) 34). R2 ) 0.9582; SEC ) 0.012 mm2/s.
Figure 4. Paired student’s t-test performed on kinematical viscosity assay results.
optimal factor
standard error of calibration (SEC) (mm2/s)
min PRESS value
2
0.012
0.0149
Table 2. Precision of the Correction Model determination times of NIRS
NIRS (mm2/s)
1 2 3 4 5 6 7
1.5865 1.5878 1.5908 1.5910 1.5917 1.5930 1.5939
ref value of ASTM445 (mm2/s)
st dev of replicates (mm2/s)
reproduction of ASTM445 (mm2/s)
1.5901
0.003
0.010
then used to find a suitable calibration. This mode of calibration is not intuitively obvious but can be very useful when fitting curves to complex data patterns. The optimal factor number of this calibration model was 2 which was the factor corresponding to the minimum prediction residual error sum of squares (PRESS) in Figure 2. Table 1 showed the parameters of the calibration model. The correlation between the reference values calculated from ASTM445 analyses and the predicted values received from the NIRS calibration model was shown in Figure 3. The variation data were good for most of the criteria under investigation. 3.2. Validation of Accuracy. The 10 samples belonging to the validation set were chosen in order to test the accuracy of the calibration model. The accuracy samples encompassed a kinematical viscosity range of 1.56-1.76 mm2/s. A comparison of the values reported by the NIR method with those of the laboratory reference method was presented in Figure 4. This difference between these two methods was statistically significant at the 95% confidence (R ) 0.05) level. This result underscored the importance of the validation of method accuracy with an independent set of samples prior to routine use of the method. The standard error of prediction (SEP) between the NIR values and the laboratory reference values was 0.027 mm2/s. 3.3. Precision of the Calibration Model. To demonstrate the repeatability of the NIR method, a single sample of aviation kerosene was analyzed seven times repeatedly. Table 2 summarized the data obtained for replicate analysis of it, along with the expected “true” value. It was evident, given the low standard deviation values, that this method performed well during replicate testing with high precision. 3.4. Influence of the Sample Residence Time in the NIR. Spectra of one sample with different sample residence times of 5, 10, and 25 min in the NIR were shown in Figure 5. Distinguished variation of the spectra, whose second derivatives curve was in the square frame, came forth due to different residence times in the dotted line square frame. The NIR prediction results of this sample with a series of sample residence times were shown in Figure 6. When the residence time of the sample in the NIR was 10 min, the prediction result was most close to the value of the reference method. The reason for this phenomenon may be that the temperature of the sample was not static due to a short residence time in the NIR or that some volatilization of the aviation kerosene occurred during a toolengthy residence time. The result of a 10 min residence time was much greater than that of some other residence times on the basis of many repeat tests. As a result, the sample should be scanned after being kept at 25 °C for 10 min.
2488 Energy & Fuels, Vol. 20, No. 6, 2006
Xing et al.
Figure 5. Spectra of one sample with different residence times (3, 10, and 25 min).
Figure 6. Effect of sample residence time in the NIR on the determination results.
4. Conclusions In this study, NIRS was successfully employed for the quantification of the kinematical viscosity of aviation kerosene
with precision similar to ASTM445. The PLS calibration model developed on 34 different aviation kerosene samples had a R2 value of 0.9582 with a standard error in the fit of 0.012. The precision of this method was estimated to be 0.003. The difference of the set of 10 accuracy samples between NIRS and ASTM445 was statistically significant at the 95% confidence level. In comparison to ASTM445 procedures, NIRS has the distinct advantages of being much faster, requiring less or no chemicals at all and no sample preparation. Thus, not only is the cost of analysis considerably reduced, but the environmental and safety concerns were also met. To improve the ruggedness of the model, a larger set of calibration samples, with a wider range in kinematical viscosity, could be employed to develop a more robust calibration model. EF060003I
