584
R. W. Kessler and W. Kessler
• The difference between cyclic-, aliphatic n-, and i-hydrocarbons is mainly due to
the number of -CH 2 -, respectively, -CH 3 in the molecule and can be differentiated
especially in the combination band range. However, at higher chain lengths, these
differences diminish.
• Aromatic hydrocarbons show the most selective spectral features in the combination band area. The aromatic bands (centered around 4600 cm
−1 ) clearly increase
with respect to the RON variation, since aromatic compounds usually have higher
RON.
• The most commonly used oxygenates like ethanol, methyl tert-butyl ether
(MTBE), and others can easily be quantified due to their high dipole moment
and show thus higher sensitivity than the hydrocarbons.
• In crude oil, bitumen, and asphaltenes measurements, the design and performance
of the sampling system are more important than the actual spectral collection.
Since these samples are very viscous and dark and have high levels of particulates,
there are high probabilities of sampling problems such as window fouling and line
plugging.
• From a more practical point of view, NIR in combination with fiber optic technology for remote sensing can be used. In this case, NIR instrumentation can be
located away from hazardous and explosive environments.
• NIR spectroscopy provides a lot faster and more repeatable data than, e.g., conventional online analysers. Furthermore, the data are based on molecular features
rather than on macroscopic parameters. The improved harmonization of real-time
analysis and process control can reach large economic benefits.
• Once correctly calibrated, NIR spectroscopy requires less maintenance than other
conventional analysers used in refineries
The use of data fusion or nonlinear chemometric tools like support vector
machines together with multiple analytical measurements clearly reduces the error
associated with inferential property models for streams with highly complex compositions, such as crude oil, its fractions, and petrochemical products. This opens the
door to a wealth of different applications and opportunities in both the laboratory
and process settings [71–73].
23.6.3.2 Example: Key Spectral Features of Gasolines: Brand Name
or Refinery?
In the previous section in Fig. 23.13, the spectral fingerprint of gasoline of different
varieties is shown. The objective of the investigation was to show, whether the differences of the varieties are due to different brands or due to the refinery structure and
the used raw material.
Figure 23.14 shows the principle component analysis (PCA) of the spectral
features in the wavelength range from 2100 nm up to 2500 nm. The results of the
other wavelength ranges are similar, but less pronounced.
R. W. Kessler and W. Kessler
• The difference between cyclic-, aliphatic n-, and i-hydrocarbons is mainly due to
the number of -CH 2 -, respectively, -CH 3 in the molecule and can be differentiated
especially in the combination band range. However, at higher chain lengths, these
differences diminish.
• Aromatic hydrocarbons show the most selective spectral features in the combination band area. The aromatic bands (centered around 4600 cm
−1 ) clearly increase
with respect to the RON variation, since aromatic compounds usually have higher
RON.
• The most commonly used oxygenates like ethanol, methyl tert-butyl ether
(MTBE), and others can easily be quantified due to their high dipole moment
and show thus higher sensitivity than the hydrocarbons.
• In crude oil, bitumen, and asphaltenes measurements, the design and performance
of the sampling system are more important than the actual spectral collection.
Since these samples are very viscous and dark and have high levels of particulates,
there are high probabilities of sampling problems such as window fouling and line
plugging.
• From a more practical point of view, NIR in combination with fiber optic technology for remote sensing can be used. In this case, NIR instrumentation can be
located away from hazardous and explosive environments.
• NIR spectroscopy provides a lot faster and more repeatable data than, e.g., conventional online analysers. Furthermore, the data are based on molecular features
rather than on macroscopic parameters. The improved harmonization of real-time
analysis and process control can reach large economic benefits.
• Once correctly calibrated, NIR spectroscopy requires less maintenance than other
conventional analysers used in refineries
The use of data fusion or nonlinear chemometric tools like support vector
machines together with multiple analytical measurements clearly reduces the error
associated with inferential property models for streams with highly complex compositions, such as crude oil, its fractions, and petrochemical products. This opens the
door to a wealth of different applications and opportunities in both the laboratory
and process settings [71–73].
23.6.3.2 Example: Key Spectral Features of Gasolines: Brand Name
or Refinery?
In the previous section in Fig. 23.13, the spectral fingerprint of gasoline of different
varieties is shown. The objective of the investigation was to show, whether the differences of the varieties are due to different brands or due to the refinery structure and
the used raw material.
Figure 23.14 shows the principle component analysis (PCA) of the spectral
features in the wavelength range from 2100 nm up to 2500 nm. The results of the
other wavelength ranges are similar, but less pronounced.