Predicting the properties of biodiesel and its blends using mid-FT-IR spectroscopy and first-order multivariate calibration

Abstract

Partial least squares regression (PLS) and support vector machine regression (SVM) were used to model the relationship between mid-FT-IR spectroscopic data and the density, refractive index and cold filter plugging point of biodiesel samples and their blends. A horizontal attenuated total reflectance mid-Fourier transform infrared spectroscopy (HATR/mid-FT-IR) method was used to measure the spectra. One hundred and forty-eight samples were prepared using biodiesel from different sources, such as canola, sunflower, corn, and soybean, along with commercial biodiesel samples purchased from a Brazilian, southern region supplier. One hundred samples were used for the calibration set, and forty-eight samples were utilized for the external validation set. The best results for predicting the cold filter plugging point were obtained using the SVM regression method, in which the root-mean-square error of prediction (RMSEP) was equal to 0.6°C. The PLS model resulted in the best prediction of the density and refractive index with RMSEP values equal to 0.2kgm−3 and 0.0001, respectively. In this work, all the biodiesel fuel properties were accurately predicted using these methodologies. Therefore, for these datasets, the PLS and SVM models demonstrated their robustness, presenting themselves as useful tools for the correlation and prediction of biodiesel properties studied using spectroscopic data.