Determination of Physicochemical Properties of Biodiesel and Blends using Low-Field NMR and Multivariate Calibration

Name: ANDRÉ FAZOLO CONSTANTINO
Type: PhD thesis
Publication date: 06/11/2018
Advisor:

Name	Role
VALDEMAR LACERDA JUNIOR	Advisor *

Examining board:

Name	Role
ÁLVARO CUNHA NETO	Internal Examiner *
DANIEL MARTELOZO CONSALTER	External Examiner *
LUCIO LEONEL BARBOSA	External Examiner *
REGINALDO BEZERRA DOS SANTOS	External Examiner *
VALDEMAR LACERDA JUNIOR	Advisor *

Summary: Because the methods specified by regulatory agencies for the determination of the physicochemical properties of biodiesel can be laborious and expensive, the development of alternative methodologies represents a major breakthrough. Thus, low-field nuclear magnetic resonance (NMR) is an advantageous option because it is nondestructive and reduces the cost and time consumption. In this study, the partial least squares (PLS) regression method was used to create models that correlated the decay curves of the Carr–Purcell–Meiboom–Gill (CPMG) signal, Continuous Wave Free Precession (CWFPx-x) or Carr–Purcell Continuous Wave Free Precession (CP-CWFPx-x), obtained from low-field NMR equipments (2.2 MHz and 15.0 MHz for 1H), with the kinematic viscosity, specific mass, refractive index and iodine value of biodiesel and their blends. Seventeen oilseeds diversified between edible and non-edible oils were utilized to synthesize the biodiesel and produce binary blends. Separately, multivariate calibration models were created only with pure biodiesel and blends with castor bean because these samples showed different tendencies from the others. The best values of root mean square error of prediction (RMSEP) for the kinematic viscosity, specific mass and refractive index were equal to 0.1 mm2/s, 1.9 kg/m3, 0.002 and 15.5 g I2/100 g of sample, respectively, for samples of biodiesel and blends without castor bean and 0.3 mm2/s, 1.3 kg/m3, 0.0003 and 1.9 g I2/100 g of sample for samples of biodiesel and blends with castor bean. The results reveal that the developed models are very satisfactory to predict the quality parameters of biodiesel and blends with fairly good efficacy, with the models created with CPMG and CP-CWFPx-x data being stood out to those constructed with CWFPx-x data. The physicochemical properties were also correlated with the decay curves of seed and oil samples, with the aim of predicting the quality of biodiesel from the analysis of its raw materials. However, the results were not very promising, since the correlations between measured physicochemical properties by American Society for Testing and Materials (ASTM) methods and its predicted values from the constructed PLS models resulted in very low coefficients of determination (R2).

Access to document