SIMULTANEOUS DETERMINATION OF QUALITY PARAMETERS IN BIODIESEL/DIESEL BLENDS USING SYNCHRONOUS FLUORESCENCE AND MULTIVARIATE ANALYSIS.

INSAUSTI MATIAS; CARLOS ROMANO; MARCELO F. PISTONESI; FERNÁNDEZ BAND BEATRIZ

MICROCHEMICAL JOURNAL

ELSEVIER SCIENCE BV

Lugar: Amsterdam; Año: 2013 vol. 108 p. 32 - 37

0026-265X

An analytical method was developed to determine four quality parameters (Biodiesel percentage, Cetane Number, Heat of Combustion Gross and Color) in biodiesel/diesel blends through a simple synchronous fluorescence spectrum of the samples. For this purpose, chemometrics models based on fluorescence spectra and PetroSpect data obtained from mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. spectrum of the samples. For this purpose, chemometrics models based on fluorescence spectra and PetroSpect data obtained from mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. fluorescence spectrum of the samples. For this purpose, chemometrics models based on fluorescence spectra and PetroSpect data obtained from mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color. fluorescence spectra and PetroSpect data obtained from mixtures of biodiesel/diesel were built. A variable selection by the successive projections algorithm (SPA) was used in order to obtain simple multiple linear regression (MLR) models based on a small subset of wavelengths. The SPA-MLR results were compared with a partial least squares (PLS) full spectrum regression. The best values found for the root mean square error of prediction using external validation were 0.37% (w/w) for the biodiesel in diesel, 0.5 for cetane number, 0.013 MJ/kg for heat of combustion and 0.1 for color.