RESUMO
Samples of distillates derived from the production of wine-fortifying spirit were analyzed for methanol by gas chromatography (GC) and near-infrared spectroscopy (NIRS). NIRS calibration models were developed which could accurately predict methanol concentrations in samples of fortifying spirit that had been produced over a period of three years from four different commercial distillation facilities. The best accuracy of the predictive models, as measured by the standard error of prediction value, was 0.06 g/L methanol. Other distillation fractions, produced during preparation of commercial fortifying spirit, were also examined. The most useful NIRS calibration models used partial least squares regression on continuous spectra from a scanning instrument, but it was demonstrated that calibrations could also be developed with a smaller number of fixed wavelengths, using multiple linear regression models. NIRS offers the advantages of rapid analysis, with simple routine operation, and may offer the potential for in-line process control in the operation of a commercial distillation facility.
