Portable near-infrared spectrometer to characterize crude oils and its
derivatives
Name: FRANCINE DALAPÍCOLA DOS SANTOS
Type: PhD thesis
Publication date: 04/04/2022
Advisor:
Name |
Role |
|---|---|
| PAULO ROBERTO FILGUEIRAS | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| ANDREAS NASCIMENTO | Internal Examiner * |
| ELCIO CRUZ DE OLIVEIRA | External Examiner * |
| EUSTAQUIO VINICIUS RIBEIRO DE CASTRO | Internal Examiner * |
| LUCIANA ASSIS TERRA | External Examiner * |
| PAULO ROBERTO FILGUEIRAS | Advisor * |
Pages
Summary: This work developed analytical methodologies based on multivariate data analysis (classification and regression) using NIR spectra obtained with a portable spectrometer. The focus was to characterize oils in terms of their physicochemical properties, classify them by origin and discriminate adulterants in oils and fuels. A portable NIR spectrometer was used and the performance of the models obtained by building the same methodology on a benchtop spectrometer was compared. The oil and fuel samples (naphtha, gasoline, diesel and kerosene) were provided by Cenpes and Transpetro, thus comprising samples from the Brazilian production field. The models used to solve the classification problems were PCA and PLS-DA. In the
construction of the methodology for petroleum, the PCA model did not allow the discrimination of the three origins of the petroleum samples, but a good response was obtained with the PLS-DA model. The performance parameters (specificity, sensitivity
and precision) in the prediction step obtained a 100% correct classification rate with the data obtained with portable equipment, while the benchtop equipment presented a classification error rate for all classes. To estimate the physicochemical properties
(ºAPI, TET, TNB, SARA) of a set of 182 oils, PLS and SVR were used. With the use of SVR, the estimation of properties with portable equipment becomes equivalent to benchtop equipment, although the portable NIR spectrometer does not acquire the spectral range between 6,000 and 4,000 cm-1, WHERE descriptive absorption bands
are located on the chemical composition. samples (combinations and first overtones of SH, NH, CH, OH bonds). Finally, in the discrimination and quantification of adulterants in oils and fuels, the PLS-DA model was able to discriminate oils from used
motor oils with sensitivity, specificity and precision above 94% and it was possible to quantify this adulterant with RMSEP of 4 .8% w/w. While for fuels, the model allowed a classification rate of 100%. While, for quantification, it was obtained for the mixtures
of gasoline and naphtha, LD is 1.3% m/m and LQ 4.4% m/m with RMSEP 1.4% m/m. Likewise, for the mixtures of diesel and kerosene, the PLS model allowed the quantification of kerosene with LD 2.8% m/m, LQ 9.3% m/m with RMSEP of 11.4% m/m. Thus, the portable spectrometer proved to be efficient and with competitive
results for all applications. It thus becomes a new tool to be explored in applications for the study of petroleum and its derivatives.
