Use of MCR-ALS with correlation constraint for NIR monitoring of soybean oil methanolysis.

Transesterification of triacylglycerol with alcohol is one of the most applied processes to biodiesel synthesis. The reaction proceeds via three reversible and consecutive steps in which monoacylglycerols and diacylglycerols are intermediates and glycerol is a byproduct. Since, in general, the methanol is added in excess related to the amount of oil, the global reaction can be simplified to one irreversible step, in which the reactants are the soybean oil (triacylglycerols) and methanol and the products are glycerol and methyl ester (biodiesel). In this work, an assessment of the use of Multivariate Curve Resolution Alternating Least Squares (MCR-ALS) with correlation constraint was carried out using a NIR in-line monitoring system. The methanolysis reactions with soybean oil (NaOH as catalyst) were carried out in a 500 mL batch system with temperature controlled by a thermostated water bath. To assess the influence of the temperature and the amount of catalyst on the reaction kinetics, batches with different temperatures (20ºC, 45°C and 55°C) and catalyst concentrations (0.75% and 1%) were produced. The methyl ester concentration profile along the reaction was estimated using MCR-ALS. Multiplicative Scatter Correction (MSC) pretreatment was employed prior to the MCR analysis. Pure spectra of each component at each temperature were used as initial estimates. Non-negativity constraint was applied in the concentration and spectral profiles. Correlation constraint in the concentration direction was applied to build models using reference values for the biodiesel content in order to recover profiles in real concentration units. Some drawbacks inherent to the system, such as rank deficiency in concentration direction, required analysis of the batches using column-wise augmented matrices composed by reactions carried out with the same system, but run under different experimental conditions. The rank deficiency was also observed in spectral direction since the triacylglycerol and the methyl ester have extremely correlated spectra. This problem motivated the use of methyl ester pure spectra to compose the column-wise augmented matrix. In addition, the fast conversion observed for the methanolysis at 55°C makes the recovery of the concentration profile more complex. Nevertheless, aside from the drawbacks mentioned, the use of MCR-ALS with a correlation constraint enabled the determination of the concentration profile of methyl ester in the methanolysis reaction. Squared correlation coefficients between the pure and resolved spectra of methyl ester were almost 100% for all models using only five reference values per analyzed batch. Comparing the experimental conditions under which the batches were produced, the concentration profiles followed the expected variation of biodiesel, allowing observation of kinetic behavior under each condition.