Search

During on-line monitoring and control of batch cooling crystallization using ATR-FTIR spectroscopy, appropriate calibration procedures are indispensable to the conversion of the spectroscopic measurements into useful and informative in-situ concentration predictions. The commonly used empirical transmittance ratio models cannot effectively handle complicated systems where the absorption bands appearing from solute and solvent overlap each other or the characteristic absorption bands of solute are rather weak and prone to noise. The conventional multivariate bilinear calibration modelling approaches of PCR and PLS are not capable of explicitly modelling with acceptable accuracy the temperature-induced spectral variations, the scaling effect resulting from the use of concentration units in mass–mass/mass–volume ratio rather than molarity, and the multiplicative effects caused by potential variations in optical path-length. In this contribution, an advanced chemometric methodology, termed extended loading space standardization, is presented and shown to be able to explicitly address the above issues. This is demonstrated by application to the on-line monitoring of the batch cooling crystallization of two organic compounds, monosodium glutamate and L-glutamic acid respectively using ATR-FTIR spectroscopy. The results showed that using extended loading space standardization has enhanced predictive performance than conventional multivariate bilinear calibration methods due to its ability in efficiently handling the impact of temperature and multiplicative effects.