Evaluation of Standardisation Methods of near Infrared Calibration Models

Abstract

The standardisation of eight partial least squares calibration models for the prediction of diesel oil properties was studied. The models were developed using spectra acquired on a laboratory accousto-optic tunable filter (AOTF) near infrared (NIR) spectrophotometer (with a quartz cuvette) and transferred to another AOTF-NIR spectrophotometer (with a fibre-optic probe) and to an Fourier transform NIR spectrometer, both designed for on-line application. Thirteen standardisation methods, using different approaches, were studied: standardisation by the pretreatment of spectra, (piecewise) direct standardisation and (piecewise) reverse standardisation. The reverse approach proved to be the best strategy to transfer the models.