Digital image-based classification of biodiesel

Energy Sources, Part A: Recovery, Utilization, and Environmental Effects, 2019

Biodiesel fuels deteriorate when exposed to air or high temperatures. In this study, biodiesel was produced through transesterification of sunflower oil and kept at 25, 45, and 70°C for 28 days. The spectrophotometry graphs of the samples were prepared and used as inputs to three classification schemes of linear discrimination analysis, principal component analysis (PCA), and weighted PCA (WPCA) to categorize the fuels into three groups, based on their acid values. The resulting average recognition accuracies were 82%, 84%, and 96% for the three classifiers, respectively. The analysis of the classifiers outputs indicated that the wavelengths between 200 and 240 nm had the higher significant role in determining the quality of the fuels. The model assessment and error analysis of the results were performed using Hotelling T 2 statistic and Q-residual index. The analysis indicated the WPCA model, at a 5% significance level, had higher power for accurately determining the quality of biodiesel samples using their spectrophotometry data.

Analytical sciences : the international journal of the Japan Society for Analytical Chemistry, 2017

A new simple, rapid and inexpensive analytical method was developed to determine the biodiesel percentage in biodiesel/diesel blends through simple digital images of samples obtained by scanning with a commercial scanner. Soybean biodiesel and petroleum diesel samples were acquired from refineries currently in operation. There were prepared several mixtures within the range 1.5 to 12.0% of biodiesel in diesel oil, using the same procedure as is done in industry. The analytical signals were images recorded with a scanner. This data was decomposed with different color systems: RGB, HSV, HLS, CMYK and Grayscale. Chemometrics models based on color signals obtained from different mixtures of biodiesel/diesel were built. The quantification by using partial least squares (PLS) resulted in a RMSEP value for biodiesel of 0.9% (w/w); this load approximately 10-times smaller than the corresponding calibration range, with a correlation of 0.96 between predicted and reference values.

Talanta, 2012

This paper investigates the use of UV-vis, near infrared (NIR) and synchronous fluorescence (SF) spectrometries coupled with multivariate classification methods to discriminate biodiesel samples with respect to the base oil employed in their production. More specifically, the present work extends previous studies by investigating the discrimination of corn-based biodiesel from two other biodiesel types (sunflower and soybean). Two classification methods are compared, namely full-spectrum SIMCA (soft independent modelling of class analogies) and SPA-LDA (linear discriminant analysis with variables selected by the successive projections algorithm). Regardless of the spectrometric technique employed, full-spectrum SIMCA did not provide an appropriate discrimination of the three biodiesel types. In contrast, all samples were correctly classified on the basis of a reduced number of wavelengths selected by SPA-LDA. It can be concluded that UV-vis, NIR and SF spectrometries can be successfully employed to discriminate corn-based biodiesel from the two other biodiesel types, but wavelength selection by SPA-LDA is key to the proper separation of the classes.

Fuel, 2019

This study addressed the detection and classification of biodiesel from different sources using electronic nose through application of statistical training-based methods and mathematical optimization techniques. Biodiesel fuels obtained from canola oil with methanol (MK), corn oil with methanol (MZ), canola oil with ethanol (EK) and corn oil with ethanol (EZ) as well as a combined fuel (EK&MZ) were mixed with different volume percentages (2, 5, 10, 20, 80 and 100) of the petroleum diesel. Data collection was performed by application of an electronic nose equipped with 8 metal oxide semiconductor (MOS) sensors. Data analysis was conducted using different methods including linear and quadratic discriminant analysis (LDA and DDA) as well as the support vector machine (SVM). Based on the results, SVM, QDA and LDA had classification precisions of 94.8%, 94.1% and 87.1%, respectively. Moreover, the discrimination and classification precision of SVM was higher (about 95.4%) for the two groups of pure and impure fuels (various mixtures of diesel and biodiesel). For QDA and LDA methods, this precision value was 84.4% and 75.5%, respectively. Classification of B5 fuels was better in all the methods when compared with B2 and B20 fuels. Detection and classification precision of B5 biodiesels was 100%, 97.6% and 96.1% for LDA, QDA and SVM methods, respectively. Application of the overall desirability function showed that QDA method had better performance when compared to LDA and SVM as it had higher discrimination and classification ability. The performance parameters of this model were 0.941, 0.941, 0.975 and 0.850 for mean precision, sensitivity, specificity and final desirability, respectively.

In this work, a multivariate approach was used to classify diesel/biodiesel fuel blends among 0% to 100% of biodiesel content on fuel mixture through discriminant analysis and cluster analysis associated with Fourier transform infrared spectroscopy (FTIR). The multivariate statistical techniques used in this work were partial least squares discriminant analysis (PLS-DA), principal component analysis (PCA), soft independent modeling of class analogy (SIMCA), hierarchical clustering analysis (HCA), and support vector machine (SVM). Multivariate analysis was performed on the following oil samples: soybean biodiesel, corn biodiesel, diesel S10, and fuel blends prepared from 0% to 100% (v/v) of biodiesel content. All multivariate statistical techniques were able to discriminate between the oil source and the ester percentage in the mixture. It was possible to develop robust multivariate models associated with the FTIR to allow for simultaneous discrimination of the types of oils used for biodiesel production and their content in fuel blends.

Journal of AOAC International, 2018

Motor oil classification is important for quality control and the identification of oil adulteration. In this work, we propose a simple, rapid, inexpensive and nondestructive approach based on image analysis and pattern recognition techniques for the classification of nine different types of motor oils according to their corresponding color histograms. For this, we applied color histogram in different color spaces such as red green blue (RGB), grayscale, and hue saturation intensity (HSI) in order to extract features that can help with the classification procedure. These color histograms and their combinations were used as input for model development and then were statistically evaluated by using linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and support vector machine (SVM) techniques. Here, two common solutions for solving a multiclass classification problem were applied: transformation to binary classification problem using a one-against-all (OAA) appr...

Chemical Engineering Communications, 2019

In this work, production of biodiesel from blends of virgin castor oil (VCO) and waste frying oil (WFO) was studied. Initial blends and the biodiesel obtained were characterized and the properties were analyzed statistically. Results indicated that the content of WFOs of the blends influenced the physical and chemical properties of the biofuel, but did not influence the rheological behavior of the biodiesel. Physical and chemical properties of biodiesel were also predicted using molecular descriptors (MDs) using several architectures of artificial neural networks (ANNs). Optimum architecture (configuration 10-2-20-3) was found for the lowest deviations in the three dependent variables (fuel acidity, ketone index, and kinematic viscosity).

Neural Computing and Applications, 2017

Four methods for direct classification (decision tree classifier, K-nearest neighbors, support vector machine and artificial neural networks) have been optimized and compared to classify biodiesel samples according to their compliance to viscosity, density, oxidative stability and iodine value, having as input the composition of fatty acid methyl esters. The eight databases for the classification of each property according to the standards EN 14214, RANP 45/2014 and ASTM D6751 were increased by synthetic minority over-sampling technique to deal with imbalanced data (different numbers of compliant and non-compliant samples). Taking as criteria the best performance and lowest complexity, decision tree classifier and K-nearest neighbors methods presented the best results, ensuring maximum accuracy, sensitivity and specificity (1.0000) for samples of external validation, for almost all properties and standards. A comparison between these methods of direct classification and empirical equations (indirect classification) distinguished positively the direct classification methods in the problem addressed.

Chemometrics and Intelligent Laboratory Systems, 2017

A multivariate approach was used for classification of fuel blends using the combined information from Fourier Transform Infrared Spectroscopy (FTIR) and stable carbon isotopes analysis by Isotope Ratio Mass Spectrometry (IRMS). Linear Discriminant Analysis (LDA) and Partial Least Squares Discriminant Analysis (PLS-DA) were applied to the classification of biodiesel/diesel fuel blends containing 0% to 100% (v/v) of biodiesel. The LDA and PLS-DA methods were able to discriminate samples ranging from 10% to 100% biodiesel (v/v) using the combined information from FTIR and IRMS. Since the global trend is toward a gradual increase in the percentage of biodiesel in fuel blends, the technique presented in this paper could be an important development in improving the traceability and identification of different raw materials used in biodiesel production.

Rona Teknik Pertanian

The use of ethanol and biodiesel, which are alternative fuels or biofuels, has increased in the last few years. Modern official standards list 25 parameters that must be determined to certify biodiesel quality. In order to determine biofuel quality, several methods were already widely used in which most of them were based on solvent extraction followed by other laboratory procedures. Yet, these methods are expensive, laborious and complicated processing for samples. Near infrared reflectance spectroscopy (NIRS) can be considered as a fast, pollution-free and non-destructive method in determining biofuel quality parameters. The objective of this study is to apply near infrared technology in classifying biodiesel based on KOH (0.3, 0.5 and 0.7) and to predict related biodiesel quality properties (water content, linolenic fatty acid, oleic acid, and stearic acid) based on its infrared reflectance. Biodiesel infrared spectrum was acquired in wavelength range from 1000 to 2500 nm for different mentioned three KOH content. Principal component analysis (PCA) with non-iterative partial least square (NIPALS) was applied to analyze biodiesel spectral data. The result showed that two principal components (PC1=97% ad PC2 = 2%) based on infrared reflectance data were successfully able to recognize and classify biodiesel based on their used KOH. Moreover, the wavelength range of 1000-1140 were to be believed related to linolenic fatty acid whilst 1450 nm and 1930 nm were associated with water content. Stearic acid can be predicted in wavelength range of 1330-1380 nm and wavelength range of 1725-1790 nm were related to oleic acid of biodiesel. This may conclude that infrared technology was feasible to use as a rapid, effective and non-invasive method in biofuel classification and evaluation.