This article was downloaded by: [McMaster University]

On: 13 March 2015, At: 08:00

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Analytical Letters
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/lanl20

Determination of Basic Nitrogen in

Residues of Crude Oil Distillation Using
ATR-FTIR and Chemometric Methods
a a b
Aline L. H. Muller , Rochele S. Picoloto , Regina C. L. Guimaraes
b b b
, Ricardo A. Guarnieri , Bianca M. S. Ferreira , Julio C. M. Dias ,
c a a
Maria F. P. Santos , Erico M. M. Flores & Edson I. Muller
a
Departamento de Química, Universidade Federal de Santa Maria ,
Campus , Santa Maria , Rio Grande do Sul , Brazil
b
Centro de Pesquisas e Desenvolvimento Leopoldo Américo Miguez
de Mello , Rio de Janeiro , Brazil
c
Departamento de Ciências Naturais , Centro Universitário Norte do
Espírito Santo , São Mateus , Espírito Santo , Brazil
Accepted author version posted online: 15 Aug 2013.Published
online: 21 Nov 2013.

To cite this article: Aline L. H. Muller , Rochele S. Picoloto , Regina C. L. Guimaraes , Ricardo A.
Guarnieri , Bianca M. S. Ferreira , Julio C. M. Dias , Maria F. P. Santos , Erico M. M. Flores & Edson I.
Muller (2013) Determination of Basic Nitrogen in Residues of Crude Oil Distillation Using ATR-FTIR and
Chemometric Methods, Analytical Letters, 46:18, 2879-2889, DOI: 10.1080/00032719.2013.828299

To link to this article: http://dx.doi.org/10.1080/00032719.2013.828299

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
 systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [McMaster University] at 08:00 13 March 2015
 Analytical Letters, 46: 2879–2889, 2013
Copyright # Taylor & Francis Group, LLC
ISSN: 0003-2719 print=1532-236X online
DOI: 10.1080/00032719.2013.828299

Chemometrics

DETERMINATION OF BASIC NITROGEN IN RESIDUES

OF CRUDE OIL DISTILLATION USING ATR-FTIR AND
CHEMOMETRIC METHODS

Aline L. H. Muller,1 Rochele S. Picoloto,1

Regina C. L. Guimaraes,2 Ricardo A. Guarnieri,2
Downloaded by [McMaster University] at 08:00 13 March 2015

Bianca M. S. Ferreira,2 Julio C. M. Dias,2

Maria F. P. Santos,3 Erico M. M. Flores,1 and
Edson I. Muller1
1
Departamento de Quı́mica, Universidade Federal de Santa Maria, Campus,
Santa Maria, Rio Grande do Sul, Brazil
2
Centro de Pesquisas e Desenvolvimento Leopoldo Américo Miguez de Mello,
Rio de Janeiro, Brazil
3
Departamento de Ciências Naturais, Centro Universitário Norte do Espı́rito
Santo, São Mateus, Espı́rito Santo, Brazil

A method for basic nitrogen determination in residues of crude oil distillation using infrared
spectroscopy and chemometrics algorithms was developed. Interval partial least squares, syn-
ergy interval partial least squares, and backward interval partial least squares were evaluated
for calibration model construction. The samples were divided into a calibration and prediction
set containing 40 and 15 samples, respectively. The first derivative with a Savitzky-Golay fil-
ter and the mean centered data showed the best results and were used in all calibration mod-
els. The backward interval partial least squares algorithm with spectra divided in 60 intervals
and combinations of 4 intervals (1407 to 1372; 1117 to 1082; 971 to 936; 914 to 879 cm
1)
showed the best root mean square error of prediction of 0.016 wt%. This calibration model
displayed a suitable correlation coefficient between reference and predicted values.

Keywords: ATR-FTIR; Backward iPLS; Basic nitrogen; Chemometrics; Crude oil residues; Interval PLS;
Synergy iPLS

INTRODUCTION
Nitrogen compounds are present in crude oil and their concentration generally
ranges from 0.1 to 0.9%. Despite this relatively low concentration, the presence of

Received 10 May 2013; accepted 25 June 2013.

The authors are grateful to INCT – Bioanalı́tica, CNPq, CAPES, and FAPERGS for supporting
this study and also to CENPES=PETROBRAS S.A. for financial support.
Address correspondence to Edson I. Muller, Departamento de Quı́mica, Universidade Federal de
Santa Maria, Campus, 97105-900, Santa Maria, Rio Grande do Sul, Brazil. E-mail: edson_muller@
yahoo.com.br

2879
 2880 A. L. H. MULLER ET AL.

nitrogen compounds can affect many of the important petroleum refining processes
(Speight 2001). Due to high polarity and basicity, nitrogen species, even at very
low concentration levels (mg kg
1), may poison the catalysts used in cracking and
hydrocracking reforming process (Mills, Boedeker, and Oblad 1950). Compounds
containing nitrogen also reduce the stability of refined products due to gum forma-
tion (Batts and Fathoni 1991). Basic and neutral nitrogen compounds can be found
in the products of the distillation of petroleum. The basic nitrogen content in crude
oil ranges from 12 to 35% of the total nitrogen, depending on boiling point of the
derivative and crude oil origin. Moreover, basic nitrogen compounds have been
found in higher boiling fractions of crude oil and do not seem to be degraded by heat-
ing process during the atmospheric and vacuum distillation (Boduszynski 1987;
Merdrignac et al. 1998; Simon et al. 2010). The main basic nitrogen compounds present
in residues of crude oil distillation were the pyridine homologues (Speight 2001).
The classical method to measure basic nitrogen in crude oil is the UOP 269-10
Downloaded by [McMaster University] at 08:00 13 March 2015

which is performed by titration in an nonaqueous system using perchloric acid diluted

in acetic acid as the titrant solution (UOP Method 269-10 2010). Determination of
basic nitrogen allows the prediction of the potential of inactivation of cracking cata-
lysts and may also be used as a control analysis for the effectiveness of hydrogenation
(Deal, Weiss, and White 1953).
Recently, Fourier transform infrared spectroscopy with attenuated total reflec-
tion (ATR-FTIR) has been used in order to minimize sample pretreatment
and increase the sample throughput for the analysis of crude oil and its fractions
(De Peinder et al. 2008; Krug, Weigel, and Kellner 1991). However, it is necessary
to make use of chemometric algorithms for analysis of complex samples spectra
obtained using infrared spectroscopy (IR) (Blanco and Bano 2003; Brereton 2000;
Rocha et al. 2012). In the literature, many papers propose the determination of
different parameters in crude oil and its derivatives using data obtained by IR and
chemometric modeling (Akhlaq 1999; Aske, Kallevik, and Sjöblom 2001; De Peinder
et al. 2009; Felkel et al. 2010; Hannisdal, Hemmingsen, and Sjöblom 2005; Jingyan,
Xiaoli, and Songbai 2012; Orrego-Ruiz et al. 2001; Satya et al. 2007; Soares, Rezende,
and Fortes 2009). Usually, partial least-squares (PLS) regression is used as multivari-
ate calibration technique for quantitative analysis and the calibration model can use
the information of the whole spectrum (Geladi and Kowalski 1986). However, in
order to improve the predictive ability of the PLS models, spectral region selection
using different algorithms have been reported in recent applications (Borin and Poppi
2005; Müller et al. 2011; Pataca et al. 2007; Silva et al. 2009; Zhou, Xiang, and Wang
2007). One of these algorithms is the interval partial least squares (iPLS), which splits
the spectrum in smaller intervals. Subsequently, models were constructed for each of
these intervals and the root mean square error of cross-validation (RMSECV) was
calculated for each model. Finally, a comparison of iPLS and RMSECV of PLS using
the full spectrum was carried out (Norgaard et al. 2000). On the other hand, synergy
interval partial least (siPLS) algorithm is carried out PLS regression models using
combination of different intervals of the spectrum (Munck et al. 2001). The last algor-
ithm is the backward interval partial least squares (biPLS) and is also considered an
extension algorithm of iPLS. In this case, PLS models were obtained and the interval
left out is one that provides the worst RMSECV and this procedure is carried out
until only one interval remains (Leardi and Norgaard 2004).
 DETERMINATION OF BASIC NITROGEN 2881

The application of variable selection methods has been performed for total acid
number (TAN) (Parisotto et al. 2010) and total sulfur determination in residues of
crude oil distillation using ATR-FTIR (Müller et al. 2012). In the first application,
the best results for TAN determination were obtained using the biPLS algorithm
with the spectrum split in twenty intervals which allowed obtaining the equivalent
root mean square error of prediction (RMSEP) in comparison with the full-spectrum
PLS model. In this case, the biPLS model used only 25% of the original variables
present in the full spectrum and it was possible to minimize the noise contribution
in the model (Parisotto et al. 2010). On the other hand, for total sulfur determi-
nation, significantly lower RMSEP was obtained using siPLS algorithm with spec-
trum divided into 20 intervals and combination of three intervals (Müller et al. 2012).
In the proposed method, the determination of basic nitrogen in residues from
crude oil distillation process was evaluated using ATR-FTIR and iPLS, siPLS,
and biPLS as algorithms for variable selection and subsequent calibration model
Downloaded by [McMaster University] at 08:00 13 March 2015

construction.

EXPERIMENTAL
Materials and Sample Preparation
Fifty five samples of crude oil distillation residues (17 and 38 residues of atmos-
pheric distillation and of vacuum distillation, respectively) with basic nitrogen values
ranging from 0.047 to 0.430 wt% were used. Basic nitrogen was determined in resi-
dues of crude oil distillation using the UOP 269-10 method (UOP Method 269-10
2010). Toluene, acetic acid, acetone, and perchloric acid of analytical grade were
purchased from Merck (Germany).
Previously, the residues of crude oil distillation samples were heated up to 180 C
to promote the melt and then the heated samples were placed on top of ATR surface
to obtain the infrared spectrum. After analysis, the ATR crystal was cleaned using
toluene in a fume-hood.

Equipment and Software

A Fourier transform infrared (FT-IR) spectrometer (Perkin Elmer, Spectrum
One, USA) provided with a universal horizontal attenuated total reflection accessory
and a ZnSe crystal with number of reflection equal to 10 (Pike Technologies, USA)
was used for sample spectra acquisition. Sample was not covered because losses of
volatile compounds were insignificant. Spectra were obtained in triplicate using a
spectral range of 4000 to 650 cm
1, resolution of 4 cm
1, and 8 scans (number of total
variables equal to 3350). Regions between 4000 to 3500 cm
1 and 2500 to 1800 cm
1
were excluded because no relevant spectral information was present (number of vari-
ables excluded equal to 1200). The spectra of two different residue of a crude oil dis-
tillation are shown in Figure 1 (black line N ¼ 0.047 wt%; and gray line ¼ 0.380 wt%)
and the regions that were excluded from calibration are indicated with rectangles. For
the calibration models, construction and interval variable selection MATLAB
version 6.5 (The Math Works, USA), using the iToolbox 2.0, was used (Norgaard
et al. 2000).
 2882 A. L. H. MULLER ET AL.
Downloaded by [McMaster University] at 08:00 13 March 2015

Figure 1. Spectra of samples with different basic nitrogen values. Black line sample with basic nitrogen ¼
0.047 wt%; and gray line sample with basic nitrogen ¼ 0.380 wt%.

Model Construction
First, the average spectrum of three replicates was obtained for each sample
and calibration models were constructed using the PLS algorithm and spectra
information without excluded regions (called all variables model, 2150 variables).
Multiplicative scatter correction (Geladi, MacDougall, and Martens 1985) (MSC)
and first derivative with a Savitzky-Golay filter (D) (Savitzky and Golay 1964) were
used as treatment. Autoscaling (A) and mean centered data (MC) were evaluated as
pre-processing algorithms for the models construction.
The suitable number of latent variables was selected using the root mean
square error of cross-validation (RMSECV). In addition, the predictive ability of
the models was compared using the root mean square error of prediction (RMSEP)
that was calculated using the samples of the predictive set. Finally, the accuracy was
evaluated using weighted least square (WLS) (González, Herador, and Asuero 1999).
For the iPLS, siPLS, and biPLS algorithms, the spectra set was split into 10,
20, 40, and 60 intervals and all possible combination of two and three intervals were
evaluated using the siPLS algorithm. Finally, biPLS models were obtained with each
interval left out.

RESULTS AND DISCUSSION

Calibration and Prediction Sample Sets
Samples were randomly divided into a calibration set to construct the models
and a prediction set which was used to calculate the RMSEP. Samples were sorted
according their respective values of basic nitrogen. Subsequently, one of every three
was selected for the prediction set. Calibration and prediction set containing 40 and
15 samples respectively were obtained. Samples with basic nitrogen values ranging
from 0.047 to 0.43 wt% and 0.061 to 0.40 wt% were selected for the calibration
and predictive set, respectively.
 DETERMINATION OF BASIC NITROGEN 2883

Spectra Features and Evaluation of Treatment and Pre-Processing

The spectra of samples with low and high basic nitrogen content obtained using
FT-IR are shown in Figure 1 and are very similar to spectra obtained in other reports
(Orrego-Ruiz et al. 2001; Jingyan et al. 2012). The main bands present in the spectra
are due to C-H stretching (3100–2500 cm
1) and C-H bending (1500–1000 cm
1).
Moreover, although the N-H stretching bands are located in the region around
3500 cm
1, significant absorbance intensities were not observed. Vibrations related
to N-H bending (1650–1550 cm
1) probably are not also observed in the spectra used
in the determination of basic nitrogen. The absence of vibrations of N-H bond can be
explained by the fact that the main basic nitrogen compounds in heavier fractions of
crude oil are pyridine homologues that do not have significant absorption bands in
these regions (Silverstein, Bassler, and Morrill 1991). Moreover, the regions between
4000 to 3500 cm
1 and 2500 to 1800 cm
1 were excluded in calibration model con-
Downloaded by [McMaster University] at 08:00 13 March 2015

struction because they showed no relevant spectral information allowing decrease

of the noise contribution in calibration models.
Multiplicative scatter correction and first derivative were evaluated as treatment
for PLS model construction. In addition, mean centered and auto-scaling data were
used as pre-processing tools. The evaluation of RMSECV values obtained for the dif-
ferent PLS models showed that there were no significant differences in these values (F
test, 95% confidence level). Moreover, the coefficient of correlation of the PLS models
were quite similar (Table 1). In order to select the best combination of treatment and
pre-processing, RMSEP values for these models were also obtained using prediction set
samples. In this sense, the lower RMSEP value was obtained using first derivative and
mean centered data. First derivative and mean centered data was also used for asphal-
tene analysis in vacuum residues using ATR-FTIR data (Orrego-Ruiz et al. 2001). In
the determination of total acid number of petroleum, the first derivative was used as the
treatment in the best model obtained because this mathematical approach minimizes
the baselines problems and improves the spectra resolution (Jingyan et al. 2012). In
order to obtain models with the best predictive ability, iPLS, siPLS, and biPLS algor-
ithm were applied using first derivative and mean centered data.

Results for iPLS Models

Different studies in the literature have shown that the use of iPLS algorithms
can improve the predictive ability of the calibration models (Borin and Poppi 2005;
Müller et al. 2011; Silva et al. 2009; Zhou et al. 2007. In Figure 2, one can observe

Table 1. Results for basic nitrogen PLS calibration models using all variables

Model VNa Intervals LVsb RMSECV (wt %) Rcal RMSEP (wt %)

PLS (MSC, A) 2150 all 4 0.047 0.940 0.056

PLS (MSC, MC) 2150 all 5 0.046 0.945 0.055
PLS (D, A) 2150 all 7 0.043 0.955 0.053
PLS (D, MC) 2150 all 6 0.048 0.966 0.050

a
VN, total variables numbers.
b
LVs, latent variables.
 2884 A. L. H. MULLER ET AL.
Downloaded by [McMaster University] at 08:00 13 March 2015

Figure 2. First derivative spectrum split in 60 different intervals. The horizontal black line represents the
RMSECV value for the full spectrum PLS model. Numbers inside the rectangles are the optimal number
of latent variables.

first derivative spectra divided in 60 different intervals. The RMSECV values of the
model with all variables are plotted in Figure 2 as a horizontal black line in order to
facilitate the comparison with RMSECV value of each interval. Models with each of
these regions were obtained and some of these models (constructed with 14, 16, and
27 interval) showed RMSECV values lower than the RMSECV value for all vari-
ables (spectra without excluded regions). However, improvement in predictive ability
was not observed for iPLS models constructed with spectra split in 10, 20, 40, and 60
intervals because significant differences were not observed in RMSEP values
obtained for this models. Probably, the spectral information correlated with the con-
tent of basic nitrogen may be located in different regions of the spectrum and iPLS
was not suitable as the interval selection algorithm.

Results for siPLS Models

The SiPLS algorithm allowed obtaining calibration models with combinations
of different spectral regions and may promote decreasing of the RMSEP values in
comparison with those obtained for PLS model with all variables. Table 2 shows
the RMSECV, RMSEP, and other information related to the performance of the
different siPLS models obtained.
In comparison between RMSECV values for siPLS models and PLS with all
variables, all models showed reduction of at least 40% in the values of RMSECV.
However, this difference is not significant when a statistical test was applied (F test,
95% confidence level). A significant decrease was also not observed for RMSEP
values obtained for siPLS models. The siPLS60 model that combined intervals
number 13 (about 2900 cm
1, probably corresponding to C-H stretching vibrations
of methyl and methylene groups present in basic nitrogen compounds) and 39 (about
 DETERMINATION OF BASIC NITROGEN 2885

Table 2. Results for basic nitrogen using siPLS calibration models

Models VNa Intervals LVsb RMSECV (wt %) Rcal RMSEP (wt %)

PLS (D, MC) 2150 all 6 0.048 0.966 0.050

siPLS10 430 6, 10 7 0.027 0.968 0.031
siPLS10 645 6, 9, 10 8 0.025 0.973 0.030
siPLS20 215 12, 20 6 0.027 0.969 0.031
siPLS20 322 12, 17, 20 8 0.025 0.974 0.029
siPLS40 108 9, 26 4 0.028 0.967 0.022
siPLS40 161 9, 16, 26 6 0.024 0.976 0.023
siPLS60 72 13, 39 4 0.027 0.970 0.021
siPLS60 108 18, 35, 39 8 0.022 0.980 0.024

a
VN, total variables numbers.
b
LVs, latent variables.
Downloaded by [McMaster University] at 08:00 13 March 2015

1300 cm
1, probably corresponding to ring stretching vibration of C-C and C-N of
the heteroaromatic nitrogen compounds) showed reduction in th RMSEP value of
about 60% in comparison with RMSEP value of PLS model (Table 2). Therefore,
the biPLS algorithm was also evaluated in order to obtain RMSEP values signifi-
cantly lower than those obtained by the PLS model with all variables.

biPLS Models
Models using the backward interval PLS algorithm also proposed the use of
spectral interval combinations to decrease the RMSEP value. Table 3 shows the
results for the best biPLS calibration models for basic nitrogen determination.
RMSECV and RMSEP values of biPLS40 and biPLS60 were significantly lower
when compared to RMSECV of PLS model with all variables (F test, 95% confidence
level). However, biPLS60 was selected as the best model for basic nitrogen determi-
nation because it was simpler and allowed a reduction of total variable numbers when
compared to the PLS model with all variables (from 2150 to 144). The biPLS60 model
was obtained when the spectra set was split into 60 intervals and the interval numbers
35, 46, 50, and 52 were used (Table 3). The selected intervals included the regions from
1407 to 1372 cm
1 (interval 35) and probably corresponds to ring stretching vibration
of C-C and C-N of the heteroaromatic nitrogen compounds. The intervals 46 (1117 to
1082 cm
1), 50 (971 to 936 cm
1), and 52 (914-879 cm
1) probably corresponded to
the bending of C-H of heteroaromatic nitrogen compounds (Silverstein et al. 1991).

Table 3. Results for basic nitrogen using biPLS calibration models

Models VNa Intervals LVsb RMSECV (wt %) Rcal RMSEP (wt %)

PLS (D, MC) 2150 all 6 0.048 0.966 0.050

biPLS10 860 5, 6, 8, 9 10 0.026 0.970 0.023
biPLS20 538 9, 12, 14, 15, 17 9 0.023 0.977 0.017
biPLS40 269 8, 14, 18, 23, 31 9 0.019 0.984 0.022
biPLS60 144 35, 46, 50, 52 9 0.017 0.990 0.016

a
VN: total variables numbers.
b
LVs: latent variables.
 2886 A. L. H. MULLER ET AL.
Downloaded by [McMaster University] at 08:00 13 March 2015

Figure 3. Errors for the prediction samples using the biPLS60 model.

The biPLS60 model showed an error lower than 10% for all prediction samples
(Figure 3).
The WLS method was also carried out with the results obtained with the
biPLS60 model for predictive sample set. The WLS results include the theoretical
expected value (1, 0), indicating that the proposed methodology is accurate and free
of systematic errors.
Figure 4 shows the values of basic N obtained for AR and VR samples obtained
by UOP 269-10 (2010) reference method and predicted values using ATR-FTIR and
the selected biPLS60 model. A correlation equal to 0.99 was obtained when the UOP
269-10 method and biPLS60 model values were compared.

Figure 4. Calibration graph obtained for the biPLS60 model using 35, 46, 50, and 52 intervals and 9 latent
variables.
 DETERMINATION OF BASIC NITROGEN 2887

CONCLUSIONS
A backward interval PLS algorithm with ATR-FTIR data provided a suitable
model for basic nitrogen determination in crude oil residues. Moreover, the RMSEP
values obtained using the best model (biPLS60) was significantly lower in compari-
son with RMSEP values of the model constructed using all variables. A relative error
lower than 10% was obtained in comparison between results of reference method
(UOP 269-10 2010) and those obtained using biPLS60 model. The proposed method
is easy to use, needs short time for analysis, uses instruments that are easy to handle,
and has low operational costs.

REFERENCES
Akhlaq, M. S. 1999. Characterization of the isolated wetting crude oil components with
Downloaded by [McMaster University] at 08:00 13 March 2015

infrared spectroscopy. J. Pet. Sci. Eng. 22: 229–235.

Aske, N., H. Kallevik, and J. Sjöblom. 2001. Determination of saturate, aromatic, resin, and
asphaltenic (SARA) components in crude oils by means of infrared and near-infrared
spectroscopy. Energy Fuels 15: 1304–1312.
Batts, B. D., and A. Z. Fathoni. 1991. A literature review on fuel stability studies with
particular emphasis on diesel oil. Energy Fuels 5: 2–21.
Blanco, M., and R. G. Bano. 2003. Determination of sugars in starch hydrolysates by IR
spectroscopy. Anal. Lett. 36: 1607–1619.
Boduszynski, M. M. 1987. Composition of heavy petroleums. 1. Molecular weight, hydrogen
deficiency, and heteroatom concentration as a function of atmospheric equivalent boiling
point up to 1400.degree.F (760.degree.C). Energy Fuels 1: 2–11.
Borin, A., and R. Poppi. 2005. Application of mid infrared spectroscopy and iPLS for the
quantification of contaminants in lubricating oil. J. Vibr. Spectrosc. 37: 27–32.
Brereton, R. G. 2000. Introduction to multivariate calibration in analytical chemistry. Analyst
125: 2125–2154.
De Peinder, P., D. D. Petrauskas, F. Singelenberg, F. Salvatori, T. Visser, F. Soulimani, and
B. M. Weckhuysen. 2008. Prediction of long and short residue properties of crude oils from
their infrared and near-infrared spectra. Appl. Spectrosc. 62: 414–422.
De Peinder, P., T. Visser, D. Petrauskas, F. Salvatori, F. Soulimani, and B. M. Weckhuysen.
2009. Prediction of long-residue properties of potential blends from mathematically mixed
infrared spectra of pure crude oils by partial least-squares regression models. Energy Fuels
23: 2164–2168.
Deal, V. Z., F. T. Weiss, and T. T. White. 1953. Determination of basic nitrogen in oils. Anal.
Chem. 3: 426–432.
Felkel, Y., N. Dörr, F. Glatz, and K. Varmuza. 2010. Determination of the total acid number
(TAN) of used gas engine oils by IR and chemometrics applying a combined strategy for
variable selection. Chemom. Intell. Lab. Syst. 101: 14–22.
Geladi, P., and B. R. Kowalski. 1986. Partial least-squares regression: A tutorial. Anal. Chim.
Acta 185: 1–17.
Geladi, P., D. MacDougall, and H. Martens. 1985. Linearization and scatter-correction for
near-infrared reflectance spectra of meat. Appl. Spectrosc. 39: 491–500.
González, A. G., M. A. Herador, and A. G. Asuero. 1999. Intra-laboratory testing of method
accuracy from recovery assays. Talanta 48: 729–736.
Hannisdal, A., P. V. Hemmingsen, and J. Sjöblom. 2005. Group-type analysis of heavy crude
oils using vibrational spectroscopy in combination with multivariate analysis. J. Ind. Eng.
Chem. Res. 44: 1349–1357.
 2888 A. L. H. MULLER ET AL.

Jingyan, L., C. Xiaoli, and T. Songbai. 2012. Research on determination of total acid number
of petroleum using mid-infrared attenuated total reflection spectroscopy. 2012. Energy Fuels
26: 5633–5637.
Krug, A., C. Weigel, and R. Kellner. 1991. Contributions to the development of an ATR-
FTIR-spectroscopic sensor for aldehydes and ketones in solutions. Anal. Lett. 24: 1257–1271.
Leardi, R., and L. Norgaard. 2004. Sequential application of backward interval partial least
squares and genetic of relevant spectral regions. J. Chemometrics 18: 486–497.
Merdrignac, I., F. Behar, P. Albrecht, P. Briot, and M. Vandenbrouke. 1998. Quantitative
extraction of nitrogen compounds in oils: Atomic balance and molecular composition. Ener.
Fuels 12: 1342–1355.
Mills, G. A., E. R. Boedeker, and A. G. Oblad. 1950. Chemical characterization of catalysts. I.
Poisoning of cracking catalysts by nitrogen compounds and potassium ion. J. Am. Chem.
Soc. 72: 1554–1560.
Müller, A. L. H., E. M. M. Flores, E. I. Müller, F. E. B. Silva, and M. F. Ferrão. 2011. Atte-
nuated total reflectance with Fourier transform infrared spectroscopy (ATR=FTIR) and
Downloaded by [McMaster University] at 08:00 13 March 2015

different PLS algorithms for simultaneous determination of clavulanic acid and amoxicillin
in powder pharmaceutical formulation. J. Braz. Chem. Soc. 22: 1903–1912.
Müller, A. L., R. S. Picoloto, P. de Azevedo Mello, M. F. Ferrão, M. de Fátima Pereira dos
Santos, R. C. Guimarães, E. I. Müller, and E. vM. Flores. 2012. Total sulfur determination
in residues of crude oil distillation using FT-IR=ATR and variable selection methods. Spec-
trochim. Acta A 89: 82–87.
Munck, L., J. P. Nielsen, B. Moller, S. Jacobsen, I. Sondergaard, S. B. Engelsen, L. Norgaard,
and R. Bro. 2001. Exploring the phenotypic expression of a regulatory proteome-altering
gene by spectroscopy and chemometrics. Anal. Chim. Acta 446: 171–186.
Norgaard, L., A. Saudland, J. Wagner, J. P. Nielsen, L. Munck, and S. B. Engelsen. 2000.
Interval partial least-squares regression (iPLS): A comparative chemometric study with
an example from near-infrared spectroscopy. Appl. Spectrosc. 54: 413–419.
Orrego-Ruiz, J. A., A. Guzmán, D. Molina, and E. Mejı́a-Ospino. 2011. Mid-infrared
attenuated total reflectance (MIR-ATR) predictive models for asphaltene contents in vac-
uum residua: asphaltene structure-functionality correlations based on partial least-squares
regression (PLS-R). Energy Fuels 25: 3678–3686.
Parisotto, G., M. F. Ferrão, A. L. H. Müller, E. I. Müller, M. F. P. Santos, R. C. L.
Guimarães, J. C. M Dias, and E. M. M. Flores. 2010. Total acid number determination
in residues of crude oil distillation using ATR-FTIR and variable selection by chemometric
methods. Energy Fuels 24: 5474–5478.
Pataca, L. M., W. B. Neto, M. C. Marcucci, and R. J. Poppi. 2007. Determination of apparent
reducing sugars, moisture and acidity in honey by attenuated total reflectance-Fourier
transform infrared spectrometry. Talanta 71: 1926–1931.
Rocha, W. F. C., B. G. Vaz, G. F. Sarmanho, L. H. C. Leal, R. Nogueira, V. F. Silva, and C.
N. Borges. 2012. Chemometrics techniques applied for classification and quantification of
binary biodiesel=diesel blends. Anal. Lett. 45: 2398–2411.
Satya, S., R. M. Roehner, M. D. Deo, F. V. Hanson. 2007. Estimation of properties of crude
oil residual fractions using chemometrics. Energy Fuels 21: 998–1005.
Savitzky, A., and M. J. E. Golay. 1964. Smoothing and differentiation of data by simplified
least squares procedures. Anal. Chem. 36: 1627–1632.
Silva, F. E. B., M. F. Ferrão, G. Parisotto, E. I. Müller, and E. M. M. Flores. 2009.
Simultaneous determination of sulphamethoxazole and trimethoprim in powder mixtures
by attenuated total reflection-Fourier transform infrared and multivariate calibration.
J. Pharm. Biomed. Anal. 49: 800–805.
Silverstein, R. M., G. C. Bassler, and T. C. Morrill. 1991. Spectrometric identification of
organic compounds. New York: John Wiley & Sons.
 DETERMINATION OF BASIC NITROGEN 2889

Simon, S., A. L. Nenniingsland, E. Herschbach, and J. Sjölom. 2010. Extraction of basic

components from petroleum crude oil. Energ. Fuels 24: 1043–1050.
Soares, I. P., T. F. Rezende, and I. C. P. Fortes. 2009. Study of the behavior changes in
physical-chemistry properties of diesel=biodiesel (B2) mixtures with residual oil and its
quantification by partial least-squares attenuated total reflection-Fourier transformed
infrared spectroscopy (PLS=ATR-FTIR). Energy Fuels 23: 4143–4148.
Speight, J. G. 2001. Handbook of petroleum analysis. New York, USA: John Wiley & Sons.
UOP Method 269–10. 2010. Nitrogen bases in hydrocarbons by titration. Conshohocken, PA:
American Society for Testing and Materials.
Zhou, X., B. Xiang, and Z. Wang. 2007. Determination of quercetin in extracts of Ginkgo
biloba L. Leaves by near-infrared reflectance spectroscopy based on interval partial least-
squares (iPLS) model. Anal. Lett. 40: 3383–3391.
Downloaded by [McMaster University] at 08:00 13 March 2015