Julio Cesar L. Alves

Biodiesel content determination in diesel fuel blends using near infrared (NIR) spectroscopy and support vector machines (SVM).

This work verifies the potential of support vector machine (SVM) algorithm applied to near infrared (NIR) spectroscopy data to develop multivariate calibration models for determination of biodiesel content in diesel fuel blends that are more effective and appropriate for analytical determinations of this type of fuel nowadays, providing the usual extended analytical range with required accuracy. Considering the difficulty to develop suitable models for this type of determination in an extended analytical range and that, in practice, biodiesel/diesel fuel blends are nowadays most often used between 0 and 30% (v/v) of biodiesel content, a calibration model is suggested for the range 0-35% (v/v) of biodiesel in diesel blends. The possibility of using a calibration model for the range 0-100% (v/v) of biodiesel in diesel fuel blends was also investigated and the difficulty in obtaining adequate results for this full analytical range is discussed. The SVM models are compared with those obtained with PLS models. The best result was obtained by the SVM model using the spectral region 4400-4600 cm(-1) providing the RMSEP value of 0.11% in 0-35% biodiesel content calibration model. This model provides the determination of biodiesel content in agreement with the accuracy required by ABNT NBR and ASTM reference methods and without interference due to the presence of vegetable oil in the mixture. The best SVM model fit performance for the relationship studied is also verified by providing similar prediction results with the use of 4400-6200 cm(-1) spectral range while the PLS results are much worse over this spectral region.

Simultaneous determination of acetylsalicylic acid, paracetamol and caffeine using solid-phase molecular fluorescence and parallel factor analysis

This paper describes the determination of acetylsalicylic acid (ASA), paracetamol and caffeine in pharmaceutical formulations using solid-phase molecular fluorescence and second order multivariate calibration. This methodology is applicable even in the presence of unknown interferences and with spectral overlap of the components in the mixture. Parallel factor analysis (PARAFAC) was used for model development, whose effectiveness was demonstrated by analysis of variance (ANOVA). Errors below 10% were obtained for all compounds using an external validation set. Benefits of the new procedures not included in the reference methods such as low cost, no need of sample preparation, simple and fast analysis using fluorescence spectrometer and no generation of waste, make this method very attractive, allowing for the simultaneous determination of compounds with good reproducibility and accuracy.

Quantification of animal fat biodiesel in soybean biodiesel and B20 diesel blends using near infrared spectroscopy and synergy interval support vector regression.

In this work, multivariate calibration based on partial least squares (PLS) and support vector regression (SVR) using the whole spectrum and variable selection by synergy interval (siPLS and siSVR) were applied to NIR spectra for the determination of animal fat biodiesel content in soybean biodiesel and B20 diesel blends. For all models, prediction errors, bias test for systematic errors and permutation test for trends in the residuals were calculated. The siSVR produced significantly lower prediction errors compared to the full spectrum methods and siPLS, with a root mean squares error (RMSEP) of 0.18%(w/w) (concentration range: 0.00%-69.00%(w/w)) in the soybean biodiesel blend and 0.10%(w/w) in the B20 diesel (concentration range: 0.00%-13.80%(w/w)). Additionally, in the models for the determination of animal fat biodiesel in blends with soybean diesel, PLS and SVR showed evidence of systematic errors, and PLS/siPLS presented trends in residuals based on the permutation test. For the B20 diesel, PLS presented evidence of systematic errors, and siPLS presented trends in the residuals.

Diesel Oil Quality Parameter Determinations Using Support Vector Regression and near Infrared Spectroscopy for Hydrotreating Feedstock Monitoring

Production monitoring and final quality control of diesel can be performed in refineries using near infrared (NIR) spectroscopy combined with regression algorithms. Partial least squares (PLS) is the multivariate regression approach commonly used for such purposes, but it is deficient for modelling complex data sets, such as found in diesel production at refineries. On the other hand, support vector regression (SVR) has demonstrated greater efficiency with high generalisation performance. The aim of this work was to develop regression models using SVR to improve the effectiveness of determining feedstock quality parameters monitored for hydrotreating process control refinery diesel production. SVR and PLS models were developed for the parameters aniline point, cetane index, density and temperature of distillation (initial boiling point and 50%, 85% and 90% recovered). The results indicate the superior modelling capability of SVR. SVR models predicted test set samples with root mean squares errors which were 21% to 54% lower than those predicted using PLS. The NIR determinations presented root mean square error lower than the reproducibility values specified by the established reference methods.

Classification of diesel pool refinery streams through near infrared spectroscopy and support vector machines using C-SVC and ν-SVC

The use of near infrared (NIR) spectroscopy combined with chemometric methods have been widely used in petroleum and petrochemical industry and provides suitable methods for process control and quality control. The algorithm support vector machines (SVM) has demonstrated to be a powerful chemometric tool for development of classification models due to its ability to nonlinear modeling and with high generalization capability and these characteristics can be especially important for treating near infrared (NIR) spectroscopy data of complex mixtures such as petroleum refinery streams. In this work, a study on the performance of the support vector machines algorithm for classification was carried out, using C-SVC and ν-SVC, applied to near infrared (NIR) spectroscopy data of different types of streams that make up the diesel pool in a petroleum refinery: light gas oil, heavy gas oil, hydrotreated diesel, kerosene, heavy naphtha and external diesel. In addition to these six streams, the diesel final blend produced in the refinery was added to complete the data set. C-SVC and ν-SVC classification models with 2, 4, 6 and 7 classes were developed for comparison between its results and also for comparison with the soft independent modeling of class analogy (SIMCA) models results. It is demonstrated the superior performance of SVC models especially using ν-SVC for development of classification models for 6 and 7 classes leading to an improvement of sensitivity on validation sample sets of 24% and 15%, respectively, when compared to SIMCA models, providing better identification of chemical compositions of different diesel pool refinery streams.

Pharmaceutical analysis in solids using front face fluorescence spectroscopy and multivariate calibration with matrix correction by piecewise direct standardization

This paper reports the application of piecewise direct standardization (PDS) for matrix correction in front face fluorescence spectroscopy of solids when different excipients are used in a pharmaceutical preparation based on a mixture of acetylsalicylic acid (ASA), paracetamol (acetaminophen) and caffeine. As verified in earlier studies, the use of different excipients and their ratio can cause a displacement, change in fluorescence intensity or band profile. To overcome this important drawback, a standardization strategy was adopted to convert all the excitation-emission fluorescence spectra into those used for model development. An excitation-emission matrix (EEM) for which excitation and emission wavelengths ranging from 265 to 405 nm and 300 to 480 nm, respectively, was used. Excellent results were obtained using unfolded partial least squares (U-PLS), with RMSEP values of 8.2 mg/g, 10.9 mg/g and 2.7 mg/g for ASA, paracetamol and caffeine, respectively, and with relative errors lesser than 5% for the three analytes.

Determining the presence of naphthenic and vegetable oils in paraffin-based lubricant oils using near infrared spectroscopy and support vector machines

The presence of naphthenic base oil and/or vegetable oil in paraffin-based lubricant oils such as automotive engine lubricant oils can severely compromise the lubricant properties and this can cause serious engine damage. A simple and fast method to identify such base stocks in mixture with paraffinic base oil and automotive engine lubricant oil is of great interest for quality monitoring. Near infrared (NIR) spectroscopy combined with chemometric methods has been applied for the development of efficient analytical methods for complex mixtures such as the ones that occur in petroleum derivatives. In this work, we carried out a study to develop classification models using support vector machines (SVM) applied to near infrared (NIR) spectroscopy data to determine the presence of naphthenic oil and/or vegetable oil in paraffin-based oils such as base oil and engine lubricant oil and the results were compared with those obtained with soft independent modeling of class analogy (SIMCA). The use of near infrared (NIR) spectroscopy and SVM provides the greatest results and a fast and simple method that achieves 95% and 100% of right predictions in the validation sample set and 87% and 75% of right predictions in the prediction sample set for the identification of naphthenic base oil, as well as for simultaneous identification of naphthenic base oil and vegetable oil in paraffin-based oils.

Near infrared spectroscopy combined with support vector machines as a process analytical chemistry tool at petroleum refineries

Introduction Process control and product quality control in petroleum refineries are often performed by means of on-line analysers which provide real-time information to a central system. This allows the adjustment of process variables if process disturbances or set point tendencies arise that differ from desired values. Analytical methods based on near infrared (NIR) spectroscopy combined with chemometric methods have been used as a tool for process analytical chemistry (PAC). This approach can provide useful information about sample composition at petroleum refineries because the various petroleum fractions and their physico-chemical characteristics are associated with different proportions of n-alkanes, iso-alkanes, cycloalkanes and aromatic compounds present in the fractions. Near infrared spectroscopy is able to distinguish between different contents of C–H bonds of methyl and methylene groups present in these compounds together with N–H, S–H and O–H bonds present in minor quantities. The partial least squares (PLS) algorithm is the one most commonly used in multivariate calibration model development. PLS is a linear method that can model some degree of non-linearity using an appropriate number of latent variables. However, certain applications of the support vector machines (SVM) algorithm have been shown to produce results better than those obtained by PLS. Feedstocks for petroleum refineries are often from different geographic regions and possess quite different characteristics. Likewise, petroleum fractions used in the production of derivatives may vary their characteristic properties over a wide range. It is difficult to include in a calibration model all the variability which may occur in practice in the refining process by means of experimental design and, moreover, complex mixtures such as petroleum fractions may not have a linear relationship with the parameter under study. For these reasons, it becomes necessary to use analytical methods that can provide non-linear calibration models with low prediction errors even for samples which have not been included in the calibration analytical range. One such method uses the SVM algorithm; it is based on statistical learning theory and can provide non-linear calibration models with high generalisation performance. These features make SVM an attractive option for the development of calibration models more effective for on-line analysis of petroleum, petroleum fractions and petroleum derivatives at petroleum refineries. The SVM algorithm was originally developed to solve pattern recognition problems and was subsequently extended to handle regression problems. Support vector machines for classification or support vector classification (SVC) simultaneously minimises the empirical error of classification and maximises the separation margin between classes using an optimal separation hyperplane, leading to a unique solution. One of the major features of SVC models is that they can operate in a kernelinduced feature space allowing non-linear modeling while good generalisation performance can be obtained even with relatively small datasets. These characteristics mean that SVC can provide a better classification performance than linear classification algorithms such as soft independent modeling of class analogy (SIMCA). To deal with regression problems, a modification of the classification algorithm was performed; in support vector regression (SVR), the use of the e-insensitive loss function that limits regression errors and penalises deviations beyond the adjusted limit during model development, helps ensure good performance. Key steps in SVC and SVR model development are the parametric optimisation and appropriate choice of the kernel function. This work reports the performance of SVM applied to NIR spectroscopy data for (1) development of regression and classification models for on-line analysis of petroleum fraction streams and diesel oil at a refinery and (2) compares the SVM predicted results with corresponding data produced by SIMCA and PLS methods.

Quantification of the contents in biojet fuel blends using near infrared spectroscopy and multivariate calibration

In this work a methodology was developed for the quantification of the contents of petroleum-derived hydroprocessed esters and fatty acids (HEFAs) and farnesane jet fuels in binary and ternary mixtures using near infrared spectroscopy and multivariate calibration based on partial least squares (PLS) regression. The developed models are simpler, faster and cheaper when compared to the standard reference method ASTM D6866-12, with a further advantage of differentiation between biofuels. Petroleum-derived jet fuel and biofuels were determined in binary blends with a root mean square error of prediction (RMSEP) value of 0.48% (v/v) for both HEFA and petroleum-derived jet fuels, with a limit of detection of 0.56% (v/v). Ternary blends presented values of RMSEP of 0.69% (v/v), 0.35% (v/v) and 0.44% (v/v) for HEFA, farnesane and petroleum-derived jet fuels, respectively, with limits of detection in the range of 0.12 to 0.24% (v/v). The developed models, based on bias and permutation tests, did not present systematic errors and trends in residuals, turning them appropriately for analysis of biojet fuel blends.

Quantification of hydrotreated vegetable oil and biodiesel contents in diesel fuel blends using near infrared spectroscopy

Introduction C urrently, the use of conventional biofuels, e.g. fatty acid methyl esters (FAME), the so-called biodiesel and paraffinic advanced biofuels, e.g. hydrotreated vegetable oil (HVO) or hydroprocessed esters and fatty acids (HEFA) known as renewable diesel, in mixtures with petroleum diesel are a feasible alternative to meet mandatory government targets to increase the use of renewable energy sources in the transport sector. Such targets exist in the European Union and United States. However, paraffinic advanced biofuels are a mixture of paraffins which are also found naturally in petroleumderived diesel. Thus, the distinction and quantification of such advanced biofuels in mixtures with biodiesel and petroleum diesel in an efficient manner is a challenge for the quality control of this type of fuel blend. Both HVO and biodiesel can be produced using feedstocks such as residual vegetable oils and animal fats. However, HVO has better quality properties, which do not vary with feedstock composition as significantly as do those of biodiesel, and are appropriate for use in low or high proportions in diesel fuel blends. HVO is the main type of paraffinic renewable diesel produced nowadays in the world but, mainly as a result of economic considerations, biodiesel is still the main substitute for petroleum diesel. HVO is a mixture of straight chain n-paraffins and i-paraffins typically with C15 to C18 carbon chains, free of aromatics, oxygen, nitrogen and sulfur, produced by catalytic hydroprocessing of triglycerides. Soybean biodiesel is mainly a mixture of methyl esters derived from triglycerides with C16 to C18 fatty acid side chains containing 0 to 3 unsaturated carbon bonds. Petroleum diesel is a mixture of mainly linear saturated hydrocarbons with C10 to C18 carbon chains, naphthenic hydrocarbons and aromatic hydrocarbons. Quantification of HVO contents in mixtures with petroleum diesel can be performed using the ASTM standard method D6866-12—“Standard test methods for determining the biobased content of solid, liquid, and gaseous samples using radiocarbon analysis”, using accelerator mass spectrometry (AMS) or liquid scintillation counter (LSC) techniques; these aim to measure the radiocarbon (C) content of a diesel fuel blend, a content that is directly related to its biofuel content. The main drawbacks of the analytical techniques applied in these methods are (1) that they do not permit the distinction and the quantification of two different biofuels contents in a mixture with petroleum diesel and (2) they are relatively expensive and time-consuming. The reference method commonly employed in the quantification of biodiesel contents in mixtures with petroleum diesel is the ASTM standard method D7371-07— “Standard test method for determination of biodiesel (fatty acid methyl esters) content in diesel fuel oil using mid infrared spectroscopy (FTIR-ATR-PLS method)”. The main drawback of this method is the use of the C=O bond stretching for calibration purposes—this approach is not selective in the presence of vegetable oil. The current study presents an innovative method to distinguish and quantify two different biofuels in diesel fuel blends. Simultaneous quantification of HVO and biodiesel in mixtures with petroleum diesel was performed using a multivariate calibration model based on near infrared (NIR) spectroscopy, providing an accurate, fast, simple and cost-effective analytical method.