|NDT.net March 2004 Vol. 9 No.03|
| 2nd MENDT Proceedings
NON-INTRUSIVE IDENTIFICATION OF THERMALLY FRACTIONATED PETROLEUM CUTSE. Hegazi
Center for Applied Physical Sciences, Research Institute,
King Fahd University of Petroleum and Minerals, Dhahran 31261,
Corresponding Author Contact:
ABSTRACTThe technique of time-resolved fluorescence spectroscopy has been applied to identify thermally fractionated petroleum cuts without resorting to sample handling. Signals collected while monitoring the laser-induced fluorescence within narrow time gates have been processed in wavelength-intensity-time domain to produce particular contour diagrams that can be used as fingerprints for the samples. The study, which used six thermally fractionated cuts of a petroleum oil sample of Super Light grade, was carried out using pulsed laser radiation of 10-ns pulse width at wavelength of 280nm and signal processing of as low as 2-ns resolution. Fingerprints of these cuts are shown along with a description of the possible photo-degradation effect on certain cuts due to the UV radiation. The results represent an interesting example of how to use this technique to monitor other petroleum products during refining.
INTRODUCTIONPetroleum oils can be characterized in different ways depending on the type of information that is sought. For instance, characterization can be done on the basis  of specific gravity, ash, salt and metal content, separation and identification of hydrocarbon group types, aliphatic and aromatic carbon and hydrogen types, etc., which all require some sort of sample preparation and the use of laboratory-resident equipment such as, hydrometers, high performance liquid chromatographs, and nuclear magnetic resonance instruments. There are cases, however, when it is important that characterization of oils be done on the spot in a non-intrusive manner without any sample preparation. This is needed in cases of inaccessible areas of chemical and thermal reactions, of real-time monitoring of production lines, and of remote oil spills. In these cases certain laser-based methods will be valuable.
There are a few laser-methods that have been employed to characterize oils, the most versatile of which is the laser-induced fluorescence method. In this case UV laser radiation is used to excite the oil sample and the resulting fluorescence signal is collected and analyzed. The different types of this fluorescence method that have been used for the oil characterization task include a) excitation and emission fluorescence [2,3] , b) synchronous and variable synchronous fluorescence [4-6 ], c) difference and derivative fluorescence [7-8] , d) contour (total luminescence) spectroscopy  , e) combination of fluorescence measurements with other techniques, such as chromatography  and cooling  , and f) time-resolved fluorescence [12-14] .
Not all of these fluorescence methods, however, can be easily applied instantly or remotely. The time-resolved fluorescence method, in particular, stands out as the most versatile in these specific applications. Customarily, this method relies on measuring the oil fluorescence decay lifetime as a function of wavelength and has been known since 1971. One of the few disadvantages of this method has to do with the need for de-convolving the fluorescence signal from the excitation laser pulse, since the oil fluorescence occurs immediately (within picoseconds) from the start of the laser pulse. Such de-convolution is not easy to do remotely, and therefore some sore of approximation will need to be employed  , which in turn will affect the actual fluorescence lifetimes.
Recently, we patented a technique to overcome this issue of de-convolution 15- [17 ]. Instead of measuring the fluorescence lifetimes as functions of emission wavelength we introduced a procedure in which the emission spectra are measured as functions of time. This is done by measuring the fluorescence spectrum at narrow gates within the temporal span of the excitation laser pulse. Therefore, for each oil sample a set of fluorescence spectra will be measured and the differences in their shapes are then highlighted in terms of contour diagrams that could serve as unique fingerprints.
In this work we apply this technique on thermally fractionated cuts of the crude oil. The objective is to introduce this method to petroleum products lighter than the non-refined crude oils, and to demonstrate the method's ability to identify, instantly and non-intrusively, chemicals and by-products during the petroleum refining processes. Six thermal fractions of one type of crude oil are used in this study, in which their time-resolved fluorescence spectra at consecutive narrow time gates will be shown along with their constructed fingerprints. It will be demonstrated that three of the fractions could not be distinguished from each other spectrally unless the time-resolved fluorescence fingerprints are compared with each other. The possible photo-degradation of one of the fractions due to excessive use of the UV laser will be also addressed.
EXPERIMENTAL SETUPThe thermally distilled fractions were prepared using ASTM D285 method  , in which 300 ml of a neat crude oil sample of Saudi Arabian Super Light grade was used. The D85 method produced fractions in the temperature ranges of 0-100, 100- 105, 105-140, 140-200, 200 ºC- end point, and the residue (corresponding to 32-212, 212-221, 221-284, 284-384, 384-end ºF point, respectively). About 3 cm3 of each fraction was then poured into a small sealed quartz cuvette, after which they were irradiated with laser pulsed radiation to excite their fluorescence spectra.
The experimental setup was a typical time-resolved laser-induced fluorescence setup, which consisted of an excitation laser and a single-channel detector system consisting of a low-resolution monochromator, a photomultiplier coupled with a gated-integrator signal processor. A schematic diagram of the setup is shown in Figure 1.
The fractionated oil samples were excited by 280-nm pulsed laser radiation from a YAG-pumped dye laser system whose final output at 560nm was frequency doubled. The resulting emission signal was digitized and sampled with the gated signal processor within 2 ns resolution. All measurements were carried out under identical controlled conditions in which signal-to-noise ratios were optimally maximized. Some samples were replenished in between runs to minimize any possible UV photo-degradation effect on the samples.
RESULTS AND DISCUSSIONFigure 2 shows the temporal profiles of the laser pulse at 280nm and the fluorescence signal at 420nm of the whole (non-fractionated) crude oil, and how the narrow time gates (TG)'s were chosen. The TG's are marked here as 0, 2, 4, 6 etc. to easily mark the left edge of each TG. For example, TG 0 is a time-gate of 2-ns width beginning exactly at the peak of the laser pulse, TG 2 is a time-gate of 2-ns width beginning 2 ns after the peak of the laser pulse, and so on. In addition to the TG's within the trailing edge of the laser pulse, we used also a TG within the leading edge, i.e., TG -2.
Figure 3 shows the laser-induced fluorescence spectra of all the thermal cuts measured with a temporal gate width of 20ns, which covers most of the lifetime of the resulting fluorescence signal, i.e., no narrow time gating was involved. It can be seen that all of these spectra, with the exception of that of the residue, have their maxima at about 360nm. The maximum of the residue fluorescence spectrum occurs at about 450nm which testifies to the fact that what remain in the residue sample are only heavy hydrocarbon compounds, since heavier aromatic compounds fluoresce at longer wavelengths than the lighter compounds. Out of the remaining five spectra in Figure 3 only two are different from the rest: the 221-284 and 284-384 ºF spectra. They are both distinct in the sense that the first has significant fluorescence intensity in the short wavelength end of the spectrum, while the latter has a sudden drop in the fluorescence intensity right after the maximum.
The other three spectra in Figure 3, namely, the 32-212, 212-221, and 384 ºF-end point cuts, all have more or less similar shapes to each other and, therefore, they cannot be distinguished from one another based on this type of information. The idea here is to use our newly presented technique to distinguish these thermal fractions from each other without having to use any chemical or physical sample preparation.
This is done by first measuring the fluorescence spectra at narrow time gates within the laser pulse time span. All samples were measured at 5 sequential 2-ns time gates, namely, TG -2, TG 0, TG 2, TG 4, and TG 6. Figure 4 shows a comparison between the fluorescence spectra that were measured at TG0 and TG6. Here the three thermally fractionated cuts whose non-gated fluorescence spectra were similar are shown in Figure 4 side by side at the bottom row, i.e., a, b, and c. In this Figure the intensities of all of the spectra at the emission wavelength of 420nm have been normalized to unity prior to plotting. This is an essential step in the fingerprinting procedure to ensure that the fingerprints are all scaled relative to one another. One can immediately see that the spectra measured at the TG0, are different in shape from those measured at TG6. This is a consequence of the fact that there are many fluorescing and non-fluorescing compounds in each sample. The fluorescing compounds all have different temporal/spectral characteristics from each other which affect the shape of the fluorescence spectrum if measured at certain narrow time windows, while the non-fluorescing compounds contribute to this shape alteration because of the quenching and energy transfer roles that they play. Therefore, the changes in the shapes of the spectra will be different from one sample to another, which makes them interesting means to use as distinguishing tools.
It is worth mentioning, before moving on, two points connected with the spectra of the 212-221, 284-384 °F thermal cuts (Figures 4b and 4e). The first is that the change in the shape of the fluorescence spectra in Figure 4e appears insignificant as compared to the other thermal cuts. This is so because the wavelength at which the normalization was made coincided with the low intensity tail of the spectrum, which is already near the baseline. If the normalization is made at a wavelength corresponding to a relatively high fluorescence the spectra would appear significantly different.
The second point is that the 212-221°F thermal cut in particular was found to be sensitive to the UV laser radiation. The fluorescence spectrum was altered rapidly as the laser was exciting it to the extent that the cuvette had to be constantly replenished with a new sample after every scan to avoid any effect due to photo-degradation. Figure 5 demonstrates this point by showing the same fluorescence spectrum before and after being photo-degraded. It can be seen that the peaks at 360nm have been totally wiped out due to photo-degradation. Figures 6a and 6b show the contour diagrams for the six thermal fractions as they were created from their corresponding normalized time-resolved fluorescence spectra. It can be seen in Figure 6a that the three previously undistinguished thermal cuts have now three different distinct fingerprints that can be used to discriminate between them. The areas to use in the discrimination is between 300-400nm. One should note here that intensity levels on all of these diagrams are relative to each other since the time-resolved spectra were all normalized in the same scale.
CONCLUSIONA newly presented approach for spectrally fingerprinting non-fractionated crude oils has been used for the first time on thermal cuts of one type of crude oil (of Super Light grade). The study was carried out using pulsed laser radiation of 10-ns pulse width at wavelength of 280nm and signal processing of as low as 2-ns resolution. Three of the thermal cuts, which were found to be indistinguishable from each other using the conventional fluorescence method, became distinguishable after constructing spectral fingerprints for them using the normalized time-resolved fluorescence spectra at 2-ns gate-widths. The results represent an interesting example of how to use this technique to monitor other petroleum products during refining.
ACKNOWLEDGEMENTThe author wishes to acknowledge the support of the Research Institute at King Fahd University of Petroleum and Minerals. He also wishes to thank Dr. M. Garwan and Dr. H. Masoudi for their support.