Near infrared spectroscopic determination of alcohols - solving non-linearity with linear and non-linear methods

The concentrations of methanol and ethanol in carbon tetrachloride solution s have been determined using near infrared (NIR) spectra in the region betw een 1100 and 2500 nm, Spectral non-linearities due to the effects of concen tration and temperature on hydrogen bonding were observed. The feasibility of obtaining accurate predictions of MeOH concentrations was assessed by ap plying multilinear regression (MLR) techniques, full spectra statistical me thods and artificial neural networks (ANN) to their NIR spectra, Principal Component Regression (PCR), Partial Least Squares (PLS) and ANN produced ca libration models with decreasing relative standard errors of prediction (RS EP) of 0.02%, 0.012% and 0.008%, respectively, compared with MLR methods ba sed on selected wavelengths which yielded a RSEP of 0.04%, Modelling of eth anol spectra gave similar results. We conclude that using full spectra chem ometric techniques and ANN methods can accurately predict alcohol concentra tions in spite of non-linearities.