Alternative butanol/gasoline and butanol/diesel fuel blends: An analysis of the interdependence between physical-chemical properties by a multivariate principal component analysis model

Abstract

The use of butanol as an oxygenated component in blends with fossil fuels has recently been recognized by the industry as a promising and green alternative for automotive use, being subject of several recent studies. In this work, the interdependence between important physical-chemical properties of butanol/gasoline and butanol/diesel fuel blends was investigated using a multivariate principal component analysis model. The model dataset was based on laboratorial results of density, kinematic viscosity, distillation, vapor pressure, octane rating, anti-knock index, flash point and cetane number in a total of 48 blends, the variables of which were transformed to principal component analysis matrix representations, pre-processed and then analyzed. A good coherence was observed between the experimental results in laboratory and those derived from the principal component analysis models, evidencing important physical-chemical changes in blends’ properties due to the butanol addition. Principal component analysis scores and loadings plots could provide an intuitive and comprehensive data visualization. Butanol/gasoline fuel blends showed an overall increase in density, octane rating and higher distillation temperatures from the initial boiling point to T60 (temperature of the 60% distilled volume) and reduction of the distillation temperatures from T70 to the final boiling point. An absolute reduction in values of all properties was observed for butanol/diesel fuel blends, especially for initial distillation temperatures from initial boiling point to T35, T98, final boiling point and flash point, whereas the reductions for density, kinematic viscosity and cetane number were less intense. Total variances of up to 92.50% and 94.14% were explained by the proposed principal component analysis model, depending on the blends matrix and butanol isomer composition.

Keywords

Blends butanol diesel gasoline PCA

Introduction

Over the past decades, different alternative liquid fuels have been investigated for complete or partial replacement of fossil fuels,¹ driven by geopolitical instabilities, oscillation in oil prices and environmental concerns.² In this context, the use of oxygenated compounds as a replacement for traditional fossil fuels has been pointed out as an elegant alternative because of their renewability and potential for reducing emissions of carbon monoxide and soot produced during the fossil fuel combustion on automotive engines.³

Historically, oxygenated compounds such as methyl and ethyl tert-butyl ether (MTBE, ETBE) or tert-amyl ether (TAME), glycol ethers, alcohols (methanol and ethanol) and fatty acids esters (biodiesel) have been the most commonly tested or employed in blends with fossil fuels. Gasoline and diesel are two classic examples of fossil fuels that have undergone continuous changes in their characteristics over time to improve their performance in automotive engines and, at the same time, meet increasingly stringent regulations around the world.

In the case of gasoline, one of the most striking transformations was associated to the octane rating. Until 1970, lead alkyl compounds were widely used to increase the anti-knock rating; however, through the 1970s they began to be discontinued due to its toxicity until they were completely banned by some countries in the mid-1990s. More recently, MTBE and mainly ethanol have been used as octane rating booster as a replacement for lead compounds, although MTBE is also being banned from some countries because of concerns over ground water contamination.⁴

From the 1990s, with the intensification of the use of ethanol in different blend proportions with gasoline, other concerns and operational problems have arisen and demanded a greater control in terms of specification, such as water content and corrosiveness (due to the high hygroscopicity of ethanol), heat of vaporization (higher for ethanol, affects the vapor pressure and is related to the cold start of engine), energy density (lower heat of combustion of ethanol), and eventual blend limitations related to high ethanol contents in gasoline, known as “blend wall”.^2,4

During the 1970s, methanol and ethanol were considered promising components for blending with diesel, but factors such as low solubility in hydrocarbons, high hygroscopicity, lower density, low cetane number (higher ignition delay) were limiting factors to the progress of these blends. Many investigations at that time were initiated to solve these problems and one option, not definitive, was the use of some additives (e.g. cetane number improvers, pour-point depressants, antismoke and antioxidants). It is also pertinent to highlight the recent worldwide efforts to minimize the sulfur content (e.g. mercaptans, sulfides, disulfides and heterocyclic) in automotive fuels, responsible for environmental problems, as well as for wear and deposits in diesel engines. Moreover, with the recent introduction of the ultra-low sulfur diesel (ULSD) more additives are being used to improve other properties as conductivity and lubricity, degraded by the reduction of the sulfur content.⁵

Butanol, a product with wide industrial application and produced by petrochemical, fermentative and catalytic routes^6–22 has recently attracted growing interest from researchers worldwide through its characteristics and potential as alternative fuel. Indeed, several papers in literature have investigated the real feasibility of using butanol as a partial surrogate of traditional petroleum derived fuels. Studies concerning not fully established fuels, such as butanol and its fuel blends, necessarily involves investigations about (1) the fuel properties and how well it meets to a specification and (2) the overall performance when applied to a given type of automotive engine.

A query into the main scientific journal databases shows that the most of recent studies have been performed regarding the effects of butanol on automotive engine performance, combustion and emissions.^23–49 However, studies focused on the evaluation of butanol fuel blends’ physical-chemical properties have still been little explored, which represents new opportunities for further research and is the motivation for the present work.

Besides, investigations concerning physical-chemical properties of not fully established fuels should consider not only the analysis of individual properties values, but also the synergetic interdependence between them, which may become a challenging task, depending on the quantity of parameters simultaneously determined. Invariably, a change in a given physical-chemical property will have influence over one or more properties, in different intensity degrees. Therefore, the understanding about the global behavior involving the fuel properties allows them to be conveniently adjusted to achieve the best quality parameters for each application. In general, this kind of analysis assumes a great complexity, not easily achieved by traditional statistical methods and, as far as our knowledge, was not yet performed for butanol blends with gasoline and diesel, as we do in this paper.

Many researchers have discovered the benefits offered by chemometrics, due to its ability to provide a better experimental model and increase the extraction of information, whether in a basic or applied research.⁵⁰ Principal component analysis (PCA) can be considered as one of the most important multivariate tools of analysis, since it is the basis of most other multivariate data analysis methods. As an exploratory tool, PCA allows to reveal the existence of anomalous samples, of relations between measured variables and of relations or groupings between samples.⁵¹ Many studies have applied the PCA to evaluate important relations between characteristics in the most varied types of fuels and biofuels.

PCA allows to summarize data with many variables in a smaller set derived from the original data, in the form of uncorrelated variables. In other words, PCA consists of decomposing a data matrix “X” containing correlated (redundant) variables into a sum of products of vectors called scores (t_h) and loadings (p_h), as shown in equation (1)

X = t_{1} p_{1}^{T} + t_{2} p_{2}^{T} + \dots . + t_{h} p_{h}^{T} + E

(1)

Where superscript “T”s represent the transpose operation, “h” is the number of principal components (PC) chosen for the model and E is the residual matrix, which contains the information not modeled in the “h” PCs.

Given that the PCs are orthogonal to each other, most of the times, it enables the reduction of the data dimensionality by eliminating its redundancy. In addition, the PCs allow an easier visualization of the systematic information present in the samples. Mathematically, the scores (t) are the projections of the samples into the space defined by the PCs, which can reveal the similarities or unexpected behavior among the samples. On the other hand, loadings value (p_1,i) is the cosine of the angle between the original variable “i” and the principal component 1 (PC1), which represent the importance of the variable “i” and PC1. The PC1 is determined following the direction of maximum variation of the data, while PC2 is orthogonal to PC1 and counts for the second direction of maximum variation of the data and so on. In general, few components are sufficient to account for a large percentage of the total variance. Residual variances are the information contained in the original set and that are not modeled by the PCs.^51–53

Honorato et al.⁵⁴ employed the PCA based on spectral data to investigate a calibration model suitable for the simultaneous determination of various gasoline properties. The scores and loadings plots allowed identifying the variables with greater weight in relation to each component and the most appropriate combination of spectral range, calibration method and type of pre-processing.

The relation between samples and variables was explored by Carvalho et al.⁵⁵ using the PCA in a study of the quality and composition of pure gasoline samples, from data obtained by gas chromatographic analysis, color, appearance, density, distillation, motor octane number, research octane number and anti-knock index. Only the first three principal components were sufficient to explain 94.7% of the total data variance (56.2% PC1, 28.4% PC2 and 10.1% PC3). The comparison of the score plots with those of the loadings, obtained from the PC1 × PC2 combination, allowed to separate 24 samples into 4 groups and to identify the most significant variables responsible for such separation, which were: temperatures of 50% and 90% distilled volume, final boiling point, isoparaffins, paraffins, naphthenes and olefins.

PCA was applied by Aleme et al.⁵⁶ to the construction of two predictive and classificatory models applied for diesel oil samples, based on their types and origins. The data were obtained from density, flash point and distillation tests in two distinct types of diesel, one with low sulfur content, for metropolitan use and another with high sulfur content for interior use. Three principal components were used, with an explained variance of 96% (64% PC1, 23.4% PC2 and 8.6% PC3) and 99.1% (61.5% PC1, 36.4% PC2 and 1.2% PC3) for each of the two models tested, respectively.

The work of Ruschel et al.⁵⁷ is an elegant example concerning the use of PCA for the classification of fuel blends. In this case, PCA along with HCA (Hierarchical Cluster Analysis) was used to classify blends of 500 ppm sulfur diesel with different types of biodiesel, using spectroscopic data obtained in an attenuated total reflectance (ATR) infrared equipment. A total variance of 99.63% was obtained using four principal components. In the proposed model, PC1 explained the increase on the biodiesel percentage in blends, PC2 differentiated blends containing methyl biodiesel from those containing ethyl, PC3 differentiated the blends containing biodiesel produced from hydrogenated vegetable fat from those produced from soybean and frying residual oils, while PC4 could separate the blends according to the type of raw material used in the biodiesel production.

In another work, PCA models based on middle infrared spectroscopy on the carbonyl group region of spectra (1700–1800 cm⁻¹) were also developed by Pimentel et al.⁵⁸ in an important study to identify raw vegetable oil contaminations and its illegal blending with petro-diesel instead of biodiesel. The PCA models generated from mid-full spectra were useful to differentiate pure petro-diesel from its blends either with biodiesel or raw oils.

In Brazil, some initiatives have currently been taken to evaluate the possibility of blending butanol with fossil fuels for internal consumption. In this sense, we have recently studied the main physical-chemical parameters of butanol/gasoline and butanol/diesel blends,^59–61 which have shown very promising results. In the present work, the interdependence between important physical-chemical properties of butanol/gasoline and butanol/diesel blends was investigated by using a PCA model and an autoscaling pre-processing technique. The different effects provided by two butanol isomers - butan-1-ol (n-butanol) and 2-methylpropan-1-ol (iso-butanol) - on blends properties were also evaluated when added in a range from 0.0% to 30.0%, in mass fraction.

Experimental section

Preparation of butanol blends

Analytical grade n-butanol and iso-butanol standards (min. 99.0%) were purchased from a local supplier in Brazil and used as received. Pure gasoline and 10 ppm sulfur content (S10) pure diesel were provided by a fuel supplier (Brasilia, DF, Brazil). A total of 48 blends were prepared varying the fuel matrix (pure gasoline or pure diesel) and the butanol isomer (n-butanol or iso-butanol).

Pre-calculated masses of butanol standards and respective fuels were individually weighed and, after that, mixed in 500 mL amber glass flasks positioned on an electronic Bel semi-analytical balance (resolution ± 0.01 g), to result in butanol contents from approximately 2.5% to 30.0%, in mass fraction. This testing interval was previously defined considering typical oxygenated concentrations adopted in Brazil (Portaria MAPA N° 75/2015),⁶² United States (ASTM D7862)⁶³ and Europe (EN 228)⁶⁴ for use in commercial blends with gasoline and diesel, thus providing better comparison parameters with these already regulated fuel blends worldwide. Blends were conveniently classified into four groups according to the matrix type, as shown in Table 1.

Table 1.

Groups of blends and respective compositions.

Group	Composition		Number of samples
Group	Hydrocarbon matrix	Butanol isomer	Number of samples
1	Gasoline	n-butanol	12
2	Gasoline	iso-butanol	12
3	Diesel	n-butanol	12
4	Diesel	iso-butanol	12
Total			48

Physical-chemical analyses

In this step, all the 48 blends and respective pure fuels mentioned in the previous experimental section “Preparation of butanol blends” were submitted to physical-chemical analyses of density (DE) and atmospheric distillation (TX, temperature of X % distilled volume). Vapor pressure (VP), octane rating (MON, RON) and anti-knock index (AKI) were analyzed only for gasoline blends, whereas kinematic viscosity (KV), flash point (FP) and cetane number (CN) only for diesel blends. Tests were performed in triplicate, in accordance with ASTM standard test methods (as shown in Table 2).

Table 2.

Fuel properties analyzed in laboratory, instrumentation and methods adopted.

Fuel property	Unit	Instrumentation	ASTM standard test method
Density (DE)	kg m⁻³	Anton Paar DMA 4500	D4052⁶⁵
Kinematic viscosity (KV)	mm² s⁻¹	Cannon CAV 2200	D445⁶⁶
Atmospheric distillation (TX)	°C	Normalab NDI 450	D86⁶⁷
Vapor pressure (VP)	kPa	Grabner Inst. MINIVAP VPS	D5191⁶⁸
Octane rating (MON), (RON), (AKI)	Octane	PetroSpec GS-1000 plus VOC	a
Flash point (FP)	°C	Tanaka atg-8 (Tag)	D56⁶⁹
Cetane number (CN)	Cetane	Adv. Eng. Tech. IQT-LM	D6890⁷⁰

^aCorrelated to ASTM D2699⁷¹ and D2700.⁷²

Density was determined by measuring the frequency of oscillation of a U-shaped glass tube filled with the sample, which was compared with the corresponding frequency of an ultrapure water, used as reference. The measured frequency was electromagnetically converted to an alternating voltage of the same frequency and related to the density. After the stabilization of the thermostatic jacket surrounding the U-tube at 20.00°C (resolution ± 0.01°C), about 3 mL of each sample was slowly injected into the DMA, 4500 densimeter, until the complete filling of the U-shaped tube, avoiding the formation of air bubbles. After each analysis, the tube was cleaned with acetone PA and homogenized with the subsequent sample.

The kinematic viscosity was determined from the time required for the sample to flow through a calibrated capillary of the CAV, 2200 viscometer. The apparatus was adjusted to operate in a constant temperature of 40°C, controlled by an external thermostatic bath, containing water and ethylene glycol at 18°C. About 10 to 20 mL of each sample was transferred into glass tubes and then placed in the sample compartment for reading, one at a time.

Until the distillation test was performed, blends were kept at 4°C to avoid evaporation losses. An exact volume of 100 mL of each blend was measured on a graduated glass cylinder and immediately transferred to a distillation flask. After emptying, the same glass cylinder was positioned at the NDI 450 condenser tube outlet. The distillation temperatures were registered by a calibrated PT100 thermocouple adjusted to the flask according to the ASTM D86. The flask containing the sample was placed in the equipment, with the base resting on a central hole ceramic plate of 38 mm or 50 mm (depending on the fuel) and the arm directed to the NDI 450 condenser tube inlet. The system was subjected to heating at a rate between 4 to 5 mL min⁻¹ and the distilled steam collected in the graduated cylinder. At the end of the distillation, the sample residue remaining in the flask was measured in a 5 mL glass beaker and the value registered on the equipment screen for correction and loss calculation. After the completion of the analysis, the distillation tube was fully cleaned to remove residues and to avoid contamination of the subsequent sample.

For the vapor pressure determination, the sample was aspirated from the flask and directly introduced into the MINIVAP VPS test chamber, which was thermostatically controlled at 100°F (37.8°C) and maintained at a vapor–liquid ratio of 4:1.

The octane rating and anti-knock index analyses in GS-1000 were performed in the medium infrared spectral region (400–4000 cm⁻¹) associated to a mathematical multiple linear regression (MLR) model. The infrared spectrum of each butanol/gasoline blend was compared to factory reference spectra contained in the GS-1000 database. The AKI for each sample was obtained from the simple arithmetic mean of their MON and RON values, automatically provided by the GS-1000. Before the analysis, it was necessary to warm-up the infrared light source for approximately 30 min. About 50 mL of each sample was transferred to a sample vial, which was coupled to the equipment arm. The pressurized system performed the fuel pumping, purging and filling of the cell with about 10 mL of the sample before the analysis was started.

The samples submitted to the flash point test were also kept at 4°C. A volume of 50 mL of each sample was measured in a graduated cylinder and transferred to a test cup of the equipment. The ignition device (pilot flame), directed to the vessel containing the sample, was started from 5°C below of the expected temperature up to the maximum limit of 10°C above, at regular intervals. Between one and another analysis, the atg-8 system was cooled with water.

For the cetane number analysis, an aliquot of approximately 100 mL of each sample was pre-filtered to remove possible impurities using a 25 mL glass syringe coupled to a 0.45 μm commercial hydrophilic hydroxyl-vinyl fluoride filter at its end. The volume of 100 mL was sufficient to perform both the purging of the system and the analysis. The sample was then transferred to the IQT fuel tank and pressurized to 50 ± 1 psi of nitrogen gas to purge the system. The combustion chamber was filled with synthetic air at a pressure of 310 ± 1 psi. Once the instrumental conditions were met, the test was started. The cetane number was calculated from the ignition delay (ms) measured by sensors that detect the interval of time between injection of the sample (injector needle lift) and the pressure increase inside the combustion chamber. The analysis sequence comprised cycles of 15 pre-injections and 32 injections, from which the corresponding mean value and standard deviation was obtained.

PCA modeling

In view of the large amount of results obtained at the physical-chemical analyses, the multivariate PCA tool was applied to interpret the experimental dataset and identify the variables with greater importance and trend in each group of samples, not always observable by the simple analysis of each isolated characteristic. Blends were compared between groups with a same fuel matrix, in the following pairs: group 1 × group 2 and group 3 × group 4 (according to the descriptions in Tables 3 and 4). For a better comparison, each pure fuel matrix was doubly associated, one with each group of blends, that is, in Table 3 two sets of triplicates (40 to 42 and 1 to 3) were assigned to the pure gasoline. The same association was done in Table 4 for the pure diesel. The experimental dataset was firstly organized in datasheets and then exported to the MatLab® 2012b licensed program for the PCA analysis, so that blends (samples) were taken as observations (lines) and properties as variables (columns) of matricial representations.

Table 3.

Numerical representation of groups 1 and 2 triplicates in the PCA score plots.

Group 1n-butanol(mass fraction)	Number in score plots	Group 2iso-butanol(mass fraction)	Number in score plots
Pure gasoline	40 to 42	Pure gasoline	1 to 3
2.55	43 to 45	2.61	4 to 6
5.17	46 to 48	5.16	7 to 9
7.87	49 to 51	7.91	10 to 12
10.67	52 to 54	10.51	13 to 15
13.42	55 to 57	13.13	16 to 18
16.11	58 to 60	15.91	19 to 21
18.83	61 to 63	18.57	22 to 24
21.46	64 to 66	21.23	25 to 27
24.08	67 to 69	23.96	28 to 30
26.90	70 to 72	26.54	31 to 33
29.52	73 to 75	29.27	34 to 36
32.14	76 to 78	31.78	37 to 39

Table 4.

Numerical representation of groups 3 and 4 triplicates in the PCA score plots.

Group 3n-butanol(mass fraction)	Number in score plots	Group 4iso-butanol(mass fraction)	Number in score plots
Pure diesel	40 to 42	Pure diesel	1 to 3
2.38	43 to 45	2.29	4 to 6
4.77	46 to 48	4.75	7 to 9
7.23	49 to 51	7.10	10 to 12
9.63	52 to 54	9.52	13 to 15
12.07	55 to 57	11.94	16 to 18
14.51	58 to 60	14.42	19 to 21
16.98	61 to 63	16.84	22 to 24
19.41	64 to 66	19.27	25 to 27
21.89	67 to 69	21.74	28 to 30
24.38	70 to 72	24.26	31 to 33
26.72	–	26.58	34 to 36
29.30	–	29.05	37 to 39

As observed in Table 4, data obtained in triplicate for two blends belonging to group 3 (73–75 and 76–78) were not included in the PCA data matrices due to inconsistencies in the distillation results for these blends, attributed to the high butanol content (above 24.38 in mass fraction). For this reason, the arrangement of data resulted in a non-symmetrical matrix between the pairs of groups 3–4.

The data were pre-processed by autoscaling, that is, mean centered and divided by the standard deviation, so that all variables have the same weight on PCA. This is necessary when different units of measurement are processed at the same time, as is the case. After entering all the data in MatLab® 2012b, the statistical PCA analysis was performed using the singular value decomposition (SVD) algorithm. The scores and loadings plots for each group of blends were generated in the software PLS-Toolbox version 7.0.3 from the Eigenvector Research Incorporated by applying the multivariate PCA analysis to the different groups tested.

Results and discussion

Physical-chemical analyses

The main results obtained for groups 1 to 4 is shown in Tables 5 and 6. For each property are reported the values for the pure gasoline and pure diesel as well as the amplitude of variation of values (the highest and the lowest) achieved in each group of blends. Distillation results are represented in Tables 5 and 6 by temperatures typically specified for gasoline (T10, T50, T90 and FBP) and diesel (T10, T50, T95).

Table 5.

Results obtained for groups 1 and 2.

Fuel property	Pure gasoline	Variation in group 1	Variation in group 2	Physical-chemical effects
DE (kg m⁻³)	728.1 ± 0.1	728.4 ± 0.2750.7 ± 0.2	728.1 ± 0.2746.9 ± 0.1	Slightly increased (up to 3.0%)
VP (kPa)	50.1 ± 0.2	48.4 ± 0.241.0 ± 0.3	48.8 ± 0.241.8 ± 0.1	Reduced (up to 18.2%)
MON (octane)	81.5 ± 0.1	81.7 ± 0.183.5 ± 0.1	82.9 ± 0.189.3 ± 0.1	Improved (up to 9.6%)
RON (octane)	89.6 ± 0.1	90.7 ± 0.196.4 ± 0.2	91.6 ± 0.095.3 ± 0.1	Improved (up to 7.6%)
AKI (octane)	85.5 ± 0.2	86.2 ± 0.189.8 ± 0.2	87.2 ± 0.191.9 ± 0.1	Improved (up to 7.5%)
Distillation (°C)
T10	58.6	59.0 – 62.2	58.5 – 63.2	Increased for
T50	95.5	92.7 – 100.6	89.2 – 96.1	T10 and T50
T90	155.2	155.9 – 122.7	155.2 – 141.9	Reduced for
FBP	202.2	202.5 – 197.4	201.5 – 195.2	T90 and FBP

DE: density; VP: vapor pressure; AKI: anti-knock index.

Table 6.

Results obtained for groups 3 and 4.

Fuel property	Pure diesel	Variation in group 3	Variation in group 4	Physical-chemical effects
DE (kg m⁻³)	834.8 ± 0.0	833.7 ± 0.1825.9 ± 0.0	833.5 ± 0.0823.4 ± 0.0	Slightly reduced (up to 1.4%)
KV (mm² s⁻¹)	2.673 ± 0.002	2.500 ± 0.0042.201 ± 0.005	2.519 ± 0.0052.305 ± 0.006	Reduced (up to 17.7%)
FP (°C)	59.0 ± 0.7	41.5 ± 0.336.0 ± 0.7	36.5 ± 0.031.5 ± 0.3	Intensely reduced (up to 46.6%)
CN (cetane)	53.3 ± 1.1	49.3 ± 0.936.8 ± 0.5	49.9 ± 1.335.7 ± 0.5	Intensely reduced (up to 33.0%)
Distillation (°C)
T10	200.8	198.2–118.7	196.4–109.1	Reduced for T10, T50 and T95
T50	265.3	264.7–236.5	263.7–234.3
T95	345.2	348.1–336.9	345.5–338.5

DE: density; KV: kinematic viscosity; FP: Flash point.

The variations in the values of properties were more intense in blends of group 1. The linear n-butanol structure favors a greater molecular packaging, incrementing density and consequently contributing to reduce the overall blends volatility. On the other hand, a further improvement of the octane rating was achieved in blends of group 2, attributed to (1) the greater difficulty of the branched molecular structure of iso-butanol to interact with oxygen molecules and (2) because of the better radical stabilization provided by the iso-butanol tertiary carbon in the combustion reaction. The results obtained for groups 1 and 2 agree with observations reported by Muzíková et al.⁷³ and Christensen et al.⁷⁴ in similar studies performed with butanol/gasoline blends.

More significant variations were identified for groups 3 and 4. Despite the slight reduction observed for DE, Table 6 shows an intense variation in viscosity, volatility and ignition quality characteristics.

The density of pure butanol is quite closer to that of pure diesel, which justifies the slight DE variation in blends. However, bold differences in their carbon chain length and boiling points were decisive to great variations in distillation temperatures. By the way, the shorter butanol carbon chain compared to those of diesel hydrocarbons reduced the overall friction among the fuel blend molecules, resulting in a greater ease to flow (lower KV). Lapuerta et al.⁷⁵ similarly reported viscosity reductions in blends prepared with diesel and C1 to C5 alcohols.

The CN reduction can be explained by the greater amount of heat captured from the IQT chamber necessary to vaporize butanol, which required a longer time for the combustion chamber to reach the ideal conditions for auto-ignition. The use of nitrates, alkyl nitrates and peroxides compounds has been proposed as cetane improvers.^76,77 Lujaji et al.⁷⁸ tested a ternary blend of 10% vegetable oil/10% n-butanol/80% diesel, which have provided CN values close to that of the pure diesel.

PCA (groups 1 and 2)

PCA modeling was performed using the first three principal components (PC1, PC2 and PC3), which explained most of the data variance (92.5%). The observation of the other PCs (PC4, PC5, etc.) only presented random behavior (not shown in Figures 1 to 4), so that the 7.5% of the variation contained in these PCs were assumed to represent only instrumental noise and random errors of the physical-chemical analyses. For a better data spatial arrangement, blends were represented by numbers, in scores plots whereas variables by acronyms, in loadings plots, as previously presented in Tables 3 and 4. Some numbers in PCA plots have been purposely omitted for an easier viewing. In this section, the weight of variables as well as the interdependence between them are presented and discussed from PCA scores and loadings plots.

The relation between butanol/gasoline blends versus PC1 scores resulted in an ascending numbering from 0.0% to 30.0%, in mass fraction, in Figure 1(a), from the most negative to the most positive PC1 value. This order, in turn, showed a direct relation with the butanol content in groups 1 and 2, which was well explained by PC1, with a variance of 67.44%. In Figure 1(b), values with the greatest deviations from the PC1 origin were attributed to the variables that had greater relation with the increase of the butanol content in groups 1 and 2. Because some variables correlated negatively with one another, PC1 axis showed a division into two quadrants, one positive and other negative.

Figure 1.

Plot of: (a) scores for groups 1 and 2 samples as a function of PC1 and (b) loadings of each variable in relation to PC1. PC: principal component; IBP: initial boiling point FBP: final boiling point; DE: density; VP: vapor pressure.

As observed in Figure 1(b), DE, MON, RON, AKI and distillation temperatures comprised from the initial boiling point (IBP) to T60 increased with the increase of the butanol content in blends, and therefore were in the positive quadrant. In an opposite way, VP, T70, T80, T90, T93, T95 and final boiling point (FBP) variables presented negative loadings, that is, their values decreased with the increase of the butanol content in blends. Within the same quadrant, we can also observe different degrees to which variables are affected, as DE, at the top of the PC1 axis (highest response), and T60, more discretely increased.

In Figure 1(b), variables were related to the PC1, generating a loading plot for groups 1 and 2. More critical properties in Figure 1(b) were located at the PC1 axis extremes and, thus, requires greater attention concerning specification limits, such as T10 and VP (the balance of which affects the engine cold start) or, also, T90 and FBP (related to other important characteristics such as performance, fuel consumption and deposits formation). The volatility behavior shown in Figure 1(b) indicates that, once blended with the gasoline, butanol molecule acts similar to medium-weight hydrocarbons with boiling points between T60 and T70, increasing distillation temperatures from IBP to T40 but reducing higher ones (from T70 to FBP).

Although not usually specified, it is worth to remember that DE has a close relationship to the fuel volatility. PC1 provides an easier visualization of the different TX tendencies observed along the loadings axis as a result of the DE increasing in blends. Despite the increase observed for MON, RON and AKI in relation to the PC1 axis, octane rating improvements provided by butanol in groups 1 and 2 were clearly limited by others, already mentioned, critical variables as T10, T90 and VP.

Figure 2(a) shows the spatial distribution of groups 1 and 2 blends in a PC2 × PC3 projection, which accounted for 21.81% and 3.25% of the total data variance, respectively. In this plot, PC2 could separate the samples into two classes (one in the PC2 positive axis and other in the PC2 negative axis), according to the butanol isomer used to produce the blend. Thus, the greater PC2 origin distances, the greater isomer distinction for a given variable. On the other hand, Figure 2(b) provides a loading representation of the groups 1 and 2 variables in a same PC2 × PC3 projection.

Figure 2.

Plot of: (a) PC2 × PC3 scores for groups 1 and 2 samples; and (b) loadings for each variable in relation to PC2 × PC3. For a better viewing, samples in (a) are only identified by the third number of the triplicate sequence presented in Table 3. DE: density; VP: vapor pressure; AKI: anti-knock index; PC: principal component; IBP: initial boiling point; FBP: final boiling point.

A simultaneous analysis of Figure 2(a) and (b) identifies the correspondent behavior of each variable according to the butanol structure. Except for some blend triplicates (4–6 and 52–54), it was noted that blends of group 1 were mostly concentrated in PC2 positive quadrant, whereas blends of group 2 in the PC2 negative quadrant. The separation between these classes in the PC2 × PC3 projection is an important indicator, because it reveals that different effects were produced by either linear or branched carbon chain structures over the blends physical-chemical properties. By the way, PC3 has a close relationship with the significance of variables, thus, the most distant variables from the PC3 origin, in Figure 2(b), had higher deviations each other for a given butanol isomer.

In Figure 2(b), T60, T70 and T50 variables presented, in this order, the highest positive values in relation to the PC2, being the most responsible for the n-butanol blends identification. Indeed, analyses showed that the central region of the distillation curve, between T50 and T70, was more or less affected depending on the butanol isomer structure, which resulted in different deviations in relation to the pure gasoline curve. At this referred distillation temperature range, the deviations for group 1 were more pronounced than for group 2. The ability of butanol to form azeotrope-like with hydrocarbons and the isomer boiling point were the main factors that defined the distillation temperatures in this range.

The range between T70 and FBP comprised precisely the transition in which the azeotrope-like mixture was distilled out from the blend, generating a strong inflection in the curve and consequently reducing distillation temperatures. The distillation behavior in Figures 1(b) and 2(b) are, therefore, a reflection of the deviation caused by the azeotropy. In the case of DE and VP, these variables were located close to the PC2 origin in Figure 2(b), but still on the positive PC2 quadrant, which can be justified by the greater packing tendency of the n-butanol linear chain compared to the branched one, resulting in a most intensely DE increasing and VP decreasing for groups 1 and 2 blends.

Conversely, MON and AKI were the variables with the greatest negative correlations with respect to the PC2, followed by T90, RON and T93 variables, that is, mostly attributed to blends produced with iso-butanol isomer. The quadrant in which the blends triplicates 34–36 and 37–39 were in Figure 2(a) was compatible with the location of the MON variable in Figure 2(b), reflecting well the significant increase in experimental octane rating tests. In the PC3 axis for group 2, MON and RON were the variables with the greatest difference each other, whereas IBP and FBP were the most different in group 1.

PCA (groups 3 and 4)

For blends of groups 3 and 4, 94.14% of the model data variance was explained by the first three principal components. The increase of the butanol content along the blends, from 0.0% to 30.0% in mass fraction, was also explained by PC1 (Figure 3(a)), with a variance of 80.83%. However, in this case, butanol caused an overall reduction of variables, so the PC1 scores were ordered from the higher to the lower values. As all variables responded positively to each other, that is, all of them were reduced with the increase of the butanol content, the PC1 axis in Figure 3(b) was not polarized.

Figure 3.

Plot of: (a) scores for groups 3 and 4 samples as a function of PC1 and (b) loadings of each variable in relation to PC1. DE: density; VP: vapor pressure; PC: principal component; IBP: initial boiling point FBP: final boiling point; FP: flash point; KV: kinematic viscosity; CN: cetane number.

Experimentally, an intense increase on the volatility was observed for all butanol/diesel blends compared to the pure diesel, so that distortions in distillation curves (azeotrope-like mixtures) were even more pronounced than those of groups 1 and 2. In the blends of groups 3 and 4, butanol was quickly evaporated during distillation tests, resulting in an intense reduction of the initial temperatures, affecting more critically the range between IBP and T35. The lower boiling point of butanol compared to those of diesel lightweight hydrocarbons also resulted in an intense FP reduction, even for blends with lower butanol content (2.5% in mass fraction).

In Figure 3(b), it is possible to emphasize the pronounced T5, FP, IBP and, mainly, T98 variables reduction for groups 3 and 4, followed by KV, T10 and FBP, which presented the lowest PC1 loadings. Despite the reduction experimentally observed for DE, T40 to T90 distillation temperatures and CN, these variables had less PC1 significance in Figure 3(b), what, according to the PCA model, indicated a greater tolerance of these variables to the butanol addition in blends, compared to other variables.

For blends of groups 3 and 4, PC2 and PC3 accounted for 9.37% and 3.94% of the total data variance, respectively (Figure 4(a)). In this case, it was also possible to observe some PC2 distinction between blends produced with either n-butanol or iso-butanol, however less evident than for groups 1 and 2. As observed in Figure 4, some blends were mistakenly located in PC2 axes, that is, some n-butanol blends were classified in the negative PC2 quadrant, whereas some iso-butanol blends in the positive PC2 quadrant. Besides that, blends triplicate 40–42, 43–45 and 49–51 were located outside of the model 95% level of confidence, explained by greater differences in values of some properties in relation to the blends set in the score plot. Based on Figure 4(b), deviation of the 40–42 triplicate can be attributed mainly to IBP and FP values, whereas deviations of 43–45 and 49–51 triplicates, mainly to higher distillation T93 and T95 ones.

Figure 4.

Plot of: (a) PC2 × PC3 scores for groups 3 and 4 samples and (b) loadings for each variable in relation to PC2 × PC3. For a better viewing, samples in (a) are only identified by the third number of the triplicate sequence presented in Table 4. DE: density; VP: vapor pressure; PC: principal component; DE: density; FBP: final boiling point; IBP: initial boiling point; KV: kinematic viscosity; CN: cetane number; FP: flash point.

This lesser PC2 distinction in Figure 4(a) indicated that the structural differences between butanol isomers (linear or branched) were not so important when blended to diesel, which may be associated to a higher solvation degree of butanol molecules by hydrocarbons of diesel. In this sense, diesel longer hydrocarbons can soften structural effects of butanol more than gasoline hydrocarbons, resulting in such PC2 differences for distinguishing isomers.

Considering PC2 × PC3 loadings in Figure 4(b), IBP and FP variables had the greatest positive contribution to the PC2, followed by T5, KV, T10, T15, DE and CN, in this order. Values significantly higher for IBP and FP were associated to blends produced with linear n-butanol, less volatile than branched iso-butanol, in the same way as with T5, T10 and T15. Most of intermediate and heavy distillation fractions from T20 to T98, and mainly FBP, had negative PC2 contribution, and then associated to blends produced with iso-butanol, so that the distinction between both isomers varied differently following the T20 to FBP direction. Variables of the group 3 were located relatively close to the PC3 origin; however, T98 and T95 had a great effect on PC3, referred to the group 4 blends.

For DE and CN variables, also in Figure 4(b), more discrete differences were observed between n-butanol and iso-butanol (variables closer to the PC2 origin). Again, DE was attributed to the positive PC2 axis because of the linear molecular packing capacity in the group 3. In the case of CN, the iso-butanol structure provided higher steric resistance for the oxygen interaction and greater radical stabilization in the tertiary carbon, making the molecule less reactive and the combustion reaction slower. So, CN is slightly favored when using the linear butanol chain, which justifies its location on the PC2 positive axis.

Table 7 summarizes the information extracted from the PCA scores and loadings plots, which shows the effects of the butanol in each group of blends tested. It is now possible to observe the good coherence of the effects provided by the proposed PCA model with those achieved in the analyses in Tables 5 and 6.

Table 7.

Summary of the butanol effects on groups 1–4 properties according to the PCA plots.

	Density (DE)	Vapor pressure (VP)	Octane MON, RON, AKI	Distillation (TX)	Kinematic viscosity (KV)	Flash point (FP)	Cetane number (CN)
Group 1	Slightly increased	Reduced	Improved	Increased from IBP to T60 and reduced from T70 to FBP	a	a	a
Group 2	Slightly increased	Reduced	Improved	Increased from IBP to T60 and reduced from T70 to FBP	a	a	a
Group 3	Slightly reduced	a	a	More intensely reduced from IBP to T35	Reduced	Intensely Reduced	Reduced
Group 4	Slightly reduced	a	a	More intensely reduced from IBP to T35	Reduced	Intensely reduced	Reduced

IBF: initial boiling point FBP: final boiling point.

^aNot analyzed.

Conclusion

In this work, the interdependence between important properties tested in butanol/gasoline and butanol/diesel blends were successfully analyzed by using PCA scores and loading plots, which resulted in intuitive and comprehensive data visualization. Furthermore, the autoscaling pre-processing of variables was especially useful for a comparison of the blends properties involving the simultaneous analysis of several variables in different units, dimensions and butanol content/type. The PCA model robustness was demonstrated from the good coherence of individual variables results compared to the experimental observations.

The projection of the variables in three principal components (PC1, PC2 and PC3) was sufficient to explain total variances of 92.50% and 94.14% for pure gasoline and diesel blends, respectively. At the scores plots, PC1 could explain the increase of the butanol content in blends and how the properties responded to it, while PC2 and PC3 differentiated between the butanol isomer in each blend and the correspondent dependence of the properties with either linear or branched isomers structures.

For PC1 plots of groups 1 and 2 (butanol/gasoline blends) an inverse relation between DE, MON, RON, AKI and distillation range from IBP to T50 was identified in relation to the VP and distillation range from T70 to FBP, according to the increase of the butanol content. PC1 and PC3 results indicated a greater association of the iso-butanol blends (group 2) with the MON and VP values, indicating a greater improvement of the octane rating and increase in volatility for the branched isomer compared to the linear one.

For groups 3 and 4 (butanol/diesel blends), an overall reduction of all properties values was evidenced by the PCA model, following the increase of the butanol content in blends. Indeed, PC1 plots showed an intense reduction of the volatility, more drastically for IBP, FP and T98, proportional to higher PC1 values. In the case of groups 3 and 4, PC2 x PC3 the separation of classes was somewhat less clear than those observed to groups 1 and 2.

The rich information here provided in the PCA scores and loadings plots can be used as shortcut to simplify future similar studies and also to predict, adjust or enhance of one or more physical-chemical properties of butanol blends, in order to meet quality and specification requirements.

Footnotes

Acknowledgements

The authors thank the fuel supplier Total S.A. for kindly providing pure gasoline and diesel samples. PAZS thanks CNPq for his research fellowship.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors thank ANP, CAPES, CNPq and FAPDF for partial financial support.

Luiz Filipe Paiva Brandão obtained his Bachelor in Chemistry degree (2005), Master in Analytical Chemistry degree (2009) and Doctorate in Analytical Chemistry (2017) at the University of Brasilia - Brazil. Since 2006, he is an official employee at the National Agency of Petroleum, Natural Gas and Biofuels, a government regulatory agency in Brazil. Since then, acts with professional activities directed to the downstream regulation of petroleum and its derivatives, natural gas and biofuels industries, such as quality control of products, inspection of laboratories/commercial agents, researches of new alternative fuels and development of new analytical methodologies applied to fuels analysis.

Jez Willian Batista Braga graduated in Chemistry from the State University of Goiás, Master and Doctorate in analytical chemistry from the State University of Campinas, postdoctoral at the Center for Nuclear Energy in Agriculture (USP-Piracicaba). Currently is a research productivity fellowship from CNPq and an Associate Professor at the Institute of Chemistry of the University of Brasilia (UnB). Present experience in teaching and research with an emphasis on chemometrics, working mainly in the following areas: Forensic Chemistry, first and second order multivariate calibration, determination of figures of merit, validation, Multivariate Curve Resolution, HPLC and spectroscopic methods (UV-VIS, IR, NIR and molecular fluorescence).

Paulo Anselmo Ziani Suarez obtained Chemical Engineer degree from Universidade Federal do Rio Grande do Sul (1993), Master in Chemistry from Universidade Federal do Rio Grande do Sul (1996) and Doctorate in Material Science from Universidade Federal do Rio Grande do Sul (2000). Has experience in Chemistry, focusing on Chemical Kinetic and Catalysis, acting on the following subjects: fossil and bio materials and fuels, multiphase catalysis and process development.

References

Garcia

Päkkillä

Ojamo

, et al. Challenges in biobutanol production: how to improve the efficiency? Renew Sust Energ Rev 2011; 15: 964–980.

Ranjan

Moholkar

VS.

Biobutanol: science, engineering, and economics. Int J Energy Res 2012; 36: 277–323.

Dogan

The influence of n-butanol/diesel fuel blends utilization on a small diesel engine performance and emissions. Fuel 2011; 90: 2467–2472.

Rand

SJ.

Significance of tests for petroleum products. 8th edn. Newburyport: ASTM International, 2010, pp.3–32.

Kozak

An application of butanol as diesel fuel component and its influence on exhaust emissions. TEKA Kom Mot i Energ Roln 2011; 11c: 126–133.

Clarke

ML.

Branched selective hydroformylation: a useful tool for organic synthesis. Curr Org Chem 2005; 9: 701–718.

Uyttebroek

Hecke

Vanbroekhoven

Sustainability metrics of 1-butanol. Catal Today 2015; 239: 7–10.

Franke

Selent

Borner

Applied hydroformylation. Chem Rev 2012; 112: 5675–5732.

Pospech

Fleischer

Franke

, et al. Alternative metals for homogeneous catalyzed hydroformylation reactions. Angew Chem Int Ed Engl 2013; 52: 2852–2872.

10.

Saha

, et al. A review of biphasic hydroformylation for long chain substrates. Res Chem Intermed 2013; 39: 3463–3474.

11.

Brito

Martins

Life cycle assessment of butanol production. Fuel 2017; 208: 476–482.

12.

Rochón

Ferrari

Lareo

Integrated ABE fermentation-gas stripping process for enhanced butanol production from sugarcane-sweet sorghum juices. Biomass Bioenergy 2017; 98: 153–160.

13.

Rubin

EM.

Genomics of cellulosic biofuels. Nat Rev 2008; 454: 841–845.

14.

Tashiro

Yoshida

Noguchi

, et al. Recent advances and future prospects for increased butanol production by acetone-butanol-ethanol fermentation. Eng Life Sci 2013; 13: 432–445.

15.

Tigunova

Shulga

Blume

YB.

Biobutanol as an alternative type of fuel. Tsitol Genet 2013; 47: 51–71.

16.

Dziegielewski

Gawron

Kazmierckzac

, et al. Butanol/biobutanol as a component of an aviation and diesel fuel. J KONES Powertrain Transport 2014; 21: 69–75.

17.

Jin

Yao

Liu

, et al. Progress in the production and application of n-butanol as a biofuel. Renew Sust Energ Rev 2011; 15: 4080–4106.

18.

Dias

MOS

Pereira

Junqueira

, et al. Butanol production in a sugarcane biorefinery using ethanol as feedstock. Part I: integration to a first generation sugarcane distillery. Chem Eng Res Des 2014; 92: 1441–1451.

19.

Dowson

GRM

Haddow

Lee

, et al. Catalytic conversion of ethanol into an advanced biofuel: unprecedented selectivity for n-butanol. Angew Chem Int Ed 2013; 52: 9005–9008.

20.

Ogo

Onda

Yanagisawa

Selective synthesis of 1-butanol from ethanol over strontium phosphate hydroxyapatite catalysts. Appl Catal A: General 2011; 402: 188–195.

21.

Sun

Wang

Recent advances in catalytic conversion of ethanol to chemicals. ACS Catal 2014; 4: 1078–1090.

22.

Riittonen

Toukoniitty

Madnani

, et al. One-pot liquid-phase catalytic conversion of ethanol to 1-butanol over aluminium oxide – the effect of the active metal on the selectivity. Catalysts 2012; 2: 68–84.

23.

Dernotte

Mounaim-Rousselle

Halter

, et al. Evaluation of butanol–gasoline blends in a port fuel-injection, spark-ignition engine. Oil Gas Sci Technol – Rev IFP 2010; 65: 345–351.

24.

Kim

Lee

Park

, et al. Engine performance and toxic gas analysis of biobutanol-blended gasoline as a vehicle fuel. Intj Automot Technol 2011; 12: 409–416.

25.

Venugopal

Ramesh

Effective utilization of butanol along with gasoline in a spark ignition engine through a dual injection system. Appl Therm Eng 2013; 59: 550–558.

26.

Masum

Masjuki

Kalam

, et al. Effect of alcohol-gasoline blends optimization on fuel properties, performance and emissions of a SI engine. J Clean Prod 2015; 86: 230–237.

27.

Yanai

Bryden

Dev

, et al. Investigation of ignition characteristics and performance of a neat n-butanol direct injection compression ignition engine at a low load. Fuel 2017; 208: 137–148.

28.

Iannuzzi

Valentino

Comparative behavior of gasoline-diesel/butanol-diesel blends and injection strategy management on performance and emissions of a light duty diesel engine. Energy 2014; 71: 321–331.

29.

Elfasakhany

Experimental study on emissions and performance of an internal combustion engine fueled with gasoline and gasoline/n-butanol blends. Energy Convers Manage 2014; 88: 277–283.

30.

Huang

Cai

, et al. Emission characteristics of a spark-ignition engine fuelled with gasoline-n-butanol blends in combination with EGR. Fuel 2012; 93: 611–617.

31.

Szwaja

Naber

JD.

Combustion of n-butanol in a spark-ignition IC engine. Fuel 2010; 89: 1573–1582.

32.

Deng

Zhang

, et al. The heat release analysis of bio-butanol/gasoline blends on a high speed SI (spark ignition) engine. Energy 2013; 60: 230–241.

33.

Yusof

MNAM

Zulkifli

NWM

Masjuki

, et al. Performance and emission characteristics of a spark ignition engine fuelled with butanol isomer-gasoline blends. Transport Res Part D 2017; 57: 23–38.

34.

Tian

Zhang

, et al. Explosion characteristics of n-butanol/iso-octane-air mixtures. Fuel 2017; 188: 90–97.

35.

Mehta

Chakraborty

Mahanta

, et al. Evaluation of fuel properties of butanol-biodiesel-diesel blends and their impact on engine performance and emissions. Ind Eng Chem Res 2010; 49: 7660–7665.

36.

Altun

Oner

Yas

, et al. Effect of n-butanol blending with a blend of diesel and biodiesel on performance and exhaust emissions of a diesel engine. Ind Eng Chem Res 2011; 50: 9425–9430.

37.

Giakoumis

Rakopoulos

Dimaratos

, et al. Exhaust emissions with ethanol or n-butanol diesel fuel blends during transient operation: a review. Renew Sust Energ Rev 2013; 17: 170–190.

38.

Chen

Liu

Han

, et al. Study on performance and emissions of a passenger-car diesel engine fueled with butanol-diesel blends. Energy 2013; 55: 638–646.

39.

Chen

Shen

Zhang

, et al. Experimental study on combustion and emission characteristics of a diesel engine fueled with 2,5-dimethylfuran-diesel, n-butanol-diesel and gasoline-diesel blends. Energy 2013; 54: 333–342.

40.

Sukjit

Herreros

Dearn

, et al. The effect of the addition of individual methyl esters on the combustion and emissions of ethanol and butanol-diesel blends. Energy 2012; 42: 364–374.

41.

Atmanlı

Ileri

Yuksel

Experimental investigation of engine performance and exhaust emissions of a diesel engine fueled with diesel-n-butanol-vegetable oil blends. Energy Convers Manage 2014; 81: 312–321.

42.

Barrios

Martín

Domínguez-Sáez

, et al. Effect of the addition of oxygenated fuels as additives on combustion characteristics and particle number and size distribution emission of a TDI diesel engine. Fuel 2014; 132: 93–100.

43.

Valentino

Corcione

Iannuzzi

, et al. Experimental study on performance and emissions of a high speed diesel engine fuelled with n-butanol diesel blends under premixed low temperature combustion. Fuel 2012; 92: 295–307.

44.

Rakopoulos

Hountalas

, et al. Investigation of the performance and emissions of bus engine operating on butanol/diesel fuel blends. Fuel 2010; 89: 2781–2790.

45.

Kumar

Saravanan

Use of higher alcohols biofuels in diesel engines: a review. Renew Sust Energ Rev 2016; 60: 84–115.

46.

Liu

Zhang

Liu

, et al. Laser diagnostics and chemical kinetic analysis of PAHs and soot in co-flow partially premixed flames using diesel surrogate and oxygenated additives of n-butanol and DMF. Combust Flame 2018; 188: 129–141.

47.

Chen

Liu

, et al. Soot reduction effects of the addition of four butanol isomers on partially premixed flames of diesel surrogates. Combust Flame 2017; 177: 123–136.

48.

Zheng

Liu

, et al. Experimental study on diesel conventional and low temperature combustion by fueling four isomers of butanol. Fuel 2015; 141: 109–119.

49.

Zheng

Yue

Liu

, et al. Effect of two-stage injection on combustion and emissions under high EGR rate on a diesel engine by fueling blends of diesel/gasoline, diesel/n-butanol, diesel/gasoline/n-butanol and pure diesel. Energ Convers Manage 2015; 90: 1–11.

50.

Lavine

Workman

JJ.

Chemometrics. Anal Chem 2013; 85: 705–714.

51.

Lyra

WS, D

Silva

de Araújo

MCU

, et al. Classificação Periódica: Um Exemplo Didático para Ensinar Análise de Componentes Principais. Quím Nova 2010; 33: 1594–1597.

52.

Jolliffe

Cadima

Principal component analysis: a review and recent developments. Philos Trans A Math Phys Eng Sci 2016; 374: 20150202

53.

Ferreira

MMC.

Quimiometria: conceitos, métodos e aplicações. 1st edn. Campinas: Unicamp, 2015, p.110.

54.

Honorato

Neto

Pimentel

, et al. Using principal component analysis to find the best calibration settings for simultaneous spectroscopic determination of several gasoline properties. Fuel 2008; 87: 3706–3709.

55.

Carvalho

FIM

Dantas

HA.

Estudo da Qualidade da Gasolina Tipo A e sua Composição Química Empregando Análise de Componentes Principais. Quím Nova 2014; 37: 33–38.

56.

Aleme

Corgozinho

CNC

Barbeira

PJS.

Diesel oil discrimination by origin and type using physicochemical properties and multivariate analysis. Fuel 2010; 89: 3151–3156.

57.

Ruschel

CFC

Huang

Samios

, et al. Análise Exploratória Aplicada a Espectros de Reflexão Total Atenuada no Infravermelho com Transformada de Fourier (ATR-FTIR) de Blendas de Biodiesel/Diesel. Quim Nova 2014; 37: 810–815.

58.

Pimentel

Ribeiro Gmgs da Cruz

, et al. Determination of biodiesel content when blended with mineral diesel fuel using infrared spectroscopy and multivariate calibration. Microchem J 2006; 82: 201–206.

59.

Brandão

LFP

Suarez

PAZ.

Determination of the alternative butanol/gasoline and butanol/diesel fuel blends heats of combustion by a heat-loss compensated semi-micro calorimeter. J Therm Anal Calorim 2018; 131: DOI 10.1007/s10973-018-7106-1.

60.

Brandão

LFP,

Suarez

PAZ.

Study of kinematic viscosity, volatility and ignition quality properties of butanol/diesel blends. Braz J Chem Eng 2018; accepted for publication.

61.

Brandão

LFP,

Suarez

PAZ.

Evaluation of the density, volatility and anti-knock rating physical-chemical properties of alternative butanol/gasoline fuel blends. Acta Sci – Technol 2017, accepted for publication.

62.

Ministério da Agricultura, Pecuária e Abastecimento (MAPA), www.agricultura.gov.br/legislacao-1 (accessed 3 March 2018).

63.

ASTM 7862:2017. Standard specification for butanol for blending with gasoline for use as automotive spark-ignition engine fuel, https://www.astm.org/Standard/standards-and-publications

64.

EN 228:2008. Automotive fuels – unleaded petrol – requirements and test methods, https://standards.cen.eu

65.

ASTM D4052:2011. Standard test method for density, relative density, and API gravity of liquids by digital density meter, https://www.astm.org/Standard/standards-and-publications

66.

ASTM D445:2012. Standard test method for kinematic viscosity of transparent and opaque liquids (and calculation of dynamic viscosity), https://www.astm.org/Standard/standards-and-publications

67.

ASTM D86:2012. Standard test method for distillation of petroleum products and liquid fuels at atmospheric pressure, https://www.astm.org/Standard/standards-and-publications

68.

ASTM D5191:2015. Standard test method for vapor pressure of petroleum products (mini method), https://www.astm.org/Standard/standards-and-publications

69.

ASTM D56:2005 (Reapproved 2010) – Standard Test Methods for Flash Point by Tag Closed Cup Tester, https://www.astm.org/Standard/standards-and-publications

70.

ASTM D6890:2015a. Standard test method for determination of ignition delay and derived cetane number (DCN) of diesel fuel oils by combustion in a constant volume chamber, https://www.astm.org/Standard/standards-and-publications

71.

ASTM 2699:2015a. Standard test method for research octane number of spark-ignition engine fuel, https://www.astm.org/Standard/standards-and-publications

72.

ASTM 2700:2016. Standard test method for motor octane number of spark-ignition engine fuel, https://www.astm.org/Standard/standards-and-publications

73.

Muzíková

ŠimáIek

Pospíšil

, et al. Density, viscosity and water phase stability of 1-butanol-gasoline blends. J Fuels 2014; 2014: 1–7. DOI: 10.1155/2014/45.

74.

Christensen

Yanowitz

Ratcliff

, et al. Renewable oxygenate blending effects on gasoline properties. Energy Fuels 2011; 25: 4723–4733.

75.

Lapuerta

García-Contreras

Campos-Fernández

, et al. Stability, lubricity, viscosity, and cold-flow properties of alcohol-diesel blends. Energy Fuels 2010; 24: 4497–4502.

76.

Liu

Influence of ethanol and cetane number (CN) improver on the ignition delay of a direct-injection diesel engine. Energy Fuels 2011; 25: 103–107.

77.

Danilov

AM.

Progress in research on fuel additives (review). Pet Chem 2015; 55: 169–179.

78.

Lujaji

Bereczky

Janosi

, et al. Cetane number and thermal properties of vegetable oil, biodiesel,1-butanol and diesel blends. J Therm Anal Calorim 2010; 102: 1175–1181.